Merge branch 'master' into rename-string

Rename all 11 nodes in the utils/string category to include a "Text"
prefix for better discoverability and natural sorting. Regex nodes get
user-friendly names without "Regex" in the display name.

Renames:
- Concatenate → Text Concatenate
- Substring → Text Substring
- Length → Text Length
- Case Converter → Text Case Converter
- Trim → Text Trim
- Replace → Text Replace
- Contains → Text Contains
- Compare → Text Compare
- Regex Match → Text Match
- Regex Extract → Text Extract Substring
- Regex Replace → Text Replace (Regex)

All renamed nodes include their old display name as a search alias so
users can still find them by searching the original name. Regex nodes
also include "regex" as a search alias.

.ci/windows_amd_base_files/run_amd_gpu_disable_smart_memory.bat

@@ -1,2 +1,2 @@
-.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --enable-dynamic-vram
+.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --disable-smart-memory
 pause

.ci/windows_intel_base_files/run_intel_gpu.bat

@@ -1,2 +0,0 @@
-.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build
-pause

.github/workflows/openapi-lint.yml

@@ -1,31 +0,0 @@
-name: OpenAPI Lint
-
-on:
-  pull_request:
-    paths:
-      - 'openapi.yaml'
-      - '.spectral.yaml'
-      - '.github/workflows/openapi-lint.yml'
-
-permissions:
-  contents: read
-
-jobs:
-  spectral:
-    name: Run Spectral
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v4
-
-      - name: Set up Node.js
-        uses: actions/setup-node@v4
-        with:
-          node-version: '20'
-
-      - name: Install Spectral
-        run: npm install -g @stoplight/spectral-cli@6
-
-      - name: Lint openapi.yaml
-        run: spectral lint openapi.yaml --ruleset .spectral.yaml --fail-severity=error

.github/workflows/release-stable-all.yml

@@ -20,12 +20,29 @@ jobs:
       git_tag: ${{ inputs.git_tag }}
       cache_tag: "cu130"
       python_minor: "13"
-      python_patch: "12"
+      python_patch: "11"
       rel_name: "nvidia"
       rel_extra_name: ""
       test_release: true
     secrets: inherit
 
+  release_nvidia_cu128:
+    permissions:
+      contents: "write"
+      packages: "write"
+      pull-requests: "read"
+    name: "Release NVIDIA cu128"
+    uses: ./.github/workflows/stable-release.yml
+    with:
+      git_tag: ${{ inputs.git_tag }}
+      cache_tag: "cu128"
+      python_minor: "12"
+      python_patch: "10"
+      rel_name: "nvidia"
+      rel_extra_name: "_cu128"
+      test_release: true
+    secrets: inherit
+
   release_nvidia_cu126:
     permissions:
       contents: "write"
@@ -59,20 +76,3 @@ jobs:
       rel_extra_name: ""
       test_release: false
     secrets: inherit
-
-  release_xpu:
-    permissions:
-      contents: "write"
-      packages: "write"
-      pull-requests: "read"
-    name: "Release Intel XPU"
-    uses: ./.github/workflows/stable-release.yml
-    with:
-      git_tag: ${{ inputs.git_tag }}
-      cache_tag: "xpu"
-      python_minor: "13"
-      python_patch: "12"
-      rel_name: "intel"
-      rel_extra_name: ""
-      test_release: true
-    secrets: inherit

.github/workflows/stable-release.yml

@@ -145,8 +145,6 @@ jobs:
           cp -r ComfyUI/.ci/windows_${{ inputs.rel_name }}_base_files/* ./
           cp ../update_comfyui_and_python_dependencies.bat ./update/
 
-          echo 'local-portable' > ComfyUI/.comfy_environment
-
           cd ..
 
           "C:\Program Files\7-Zip\7z.exe" a -t7z -m0=lzma2 -mx=9 -mfb=128 -md=768m -ms=on -mf=BCJ2 ComfyUI_windows_portable.7z ComfyUI_windows_portable

.github/workflows/tag-dispatch-cloud.yml

@@ -1,45 +0,0 @@
-name: Tag Dispatch to Cloud
-
-on:
-  push:
-    tags:
-      - 'v*'
-
-jobs:
-  dispatch-cloud:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Send repository dispatch to cloud
-        env:
-          DISPATCH_TOKEN: ${{ secrets.CLOUD_REPO_DISPATCH_TOKEN }}
-          RELEASE_TAG: ${{ github.ref_name }}
-        run: |
-          set -euo pipefail
-
-          if [ -z "${DISPATCH_TOKEN:-}" ]; then
-            echo "::error::CLOUD_REPO_DISPATCH_TOKEN is required but not set."
-            exit 1
-          fi
-
-          RELEASE_URL="https://github.com/${{ github.repository }}/releases/tag/${RELEASE_TAG}"
-
-          PAYLOAD="$(jq -n \
-            --arg release_tag "$RELEASE_TAG" \
-            --arg release_url "$RELEASE_URL" \
-            '{
-              event_type: "comfyui_tag_pushed",
-              client_payload: {
-                release_tag: $release_tag,
-                release_url: $release_url
-              }
-            }')"
-
-          curl -fsSL \
-            -X POST \
-            -H "Accept: application/vnd.github+json" \
-            -H "Content-Type: application/json" \
-            -H "Authorization: Bearer ${DISPATCH_TOKEN}" \
-            https://api.github.com/repos/Comfy-Org/cloud/dispatches \
-            -d "$PAYLOAD"
-
-          echo "✅ Dispatched ComfyUI tag ${RELEASE_TAG} to Comfy-Org/cloud"

.gitignore

@@ -21,6 +21,6 @@ venv*/
 *.log
 web_custom_versions/
 .DS_Store
+openapi.yaml
 filtered-openapi.yaml
 uv.lock
-.comfy_environment

.spectral.yaml

@@ -1,91 +0,0 @@
-extends:
-  - spectral:oas
-
-# Severity levels: error, warn, info, hint, off
-# Rules from the built-in "spectral:oas" ruleset are active by default.
-# Below we tune severity and add custom rules for our conventions.
-#
-# This ruleset mirrors Comfy-Org/cloud/.spectral.yaml so specs across the
-# organization are linted against a single consistent standard.
-
-rules:
-  # -----------------------------------------------------------------------
-  # Built-in rule severity overrides
-  # -----------------------------------------------------------------------
-  operation-operationId: error
-  operation-description: warn
-  operation-tag-defined: error
-  info-contact: off
-  info-description: warn
-  no-eval-in-markdown: error
-  no-$ref-siblings: error
-
-  # -----------------------------------------------------------------------
-  # Custom rules: naming conventions
-  # -----------------------------------------------------------------------
-
-  # Property names should be snake_case
-  property-name-snake-case:
-    description: Property names must be snake_case
-    severity: warn
-    given: "$.components.schemas.*.properties[*]~"
-    then:
-      function: pattern
-      functionOptions:
-        match: "^[a-z][a-z0-9]*(_[a-z0-9]+)*$"
-
-  # Operation IDs should be camelCase
-  operation-id-camel-case:
-    description: Operation IDs must be camelCase
-    severity: warn
-    given: "$.paths.*.*.operationId"
-    then:
-      function: pattern
-      functionOptions:
-        match: "^[a-z][a-zA-Z0-9]*$"
-
-  # -----------------------------------------------------------------------
-  # Custom rules: response conventions
-  # -----------------------------------------------------------------------
-
-  # Error responses (4xx, 5xx) should use a consistent shape
-  error-response-schema:
-    description: Error responses should reference a standard error schema
-    severity: hint
-    given: "$.paths.*.*.responses[?(@property >= '400' && @property < '600')].content['application/json'].schema"
-    then:
-      field: "$ref"
-      function: truthy
-
-  # All 2xx responses with JSON body should have a schema
-  response-schema-defined:
-    description: Success responses with JSON content should define a schema
-    severity: warn
-    given: "$.paths.*.*.responses[?(@property >= '200' && @property < '300')].content['application/json']"
-    then:
-      field: schema
-      function: truthy
-
-  # -----------------------------------------------------------------------
-  # Custom rules: best practices
-  # -----------------------------------------------------------------------
-
-  # Path parameters must have a description
-  path-param-description:
-    description: Path parameters should have a description
-    severity: warn
-    given:
-      - "$.paths.*.parameters[?(@.in == 'path')]"
-      - "$.paths.*.*.parameters[?(@.in == 'path')]"
-    then:
-      field: description
-      function: truthy
-
-  # Schemas should have a description
-  schema-description:
-    description: Component schemas should have a description
-    severity: hint
-    given: "$.components.schemas.*"
-    then:
-      field: description
-      function: truthy

CODEOWNERS

@@ -1,2 +1,2 @@
 # Admins
-* @comfyanonymous @kosinkadink @guill @alexisrolland @rattus128 @kijai
+* @comfyanonymous @kosinkadink @guill

QUANTIZATION.md

@@ -139,9 +139,9 @@ Example:
   "_quantization_metadata": {
     "format_version": "1.0",
     "layers": {
-      "model.layers.0.mlp.up_proj": {"format": "float8_e4m3fn"},
-      "model.layers.0.mlp.down_proj": {"format": "float8_e4m3fn"},
-      "model.layers.1.mlp.up_proj": {"format": "float8_e4m3fn"}
+      "model.layers.0.mlp.up_proj": "float8_e4m3fn",
+      "model.layers.0.mlp.down_proj": "float8_e4m3fn",
+      "model.layers.1.mlp.up_proj": "float8_e4m3fn"
     }
   }
 }
@@ -165,4 +165,4 @@ Activation quantization (e.g., for FP8 Tensor Core operations) requires `input_s
 3. **Compute scales**: Derive `input_scale` from collected statistics
 4. **Store in checkpoint**: Save `input_scale` parameters alongside weights
 
-The calibration dataset should be representative of your target use case. For diffusion models, this typically means a diverse set of prompts and generation parameters.
+The calibration dataset should be representative of your target use case. For diffusion models, this typically means a diverse set of prompts and generation parameters.

README.md

@@ -1,7 +1,7 @@
 <div align="center">
 
 # ComfyUI
-**The most powerful and modular AI engine for content creation.**
+**The most powerful and modular visual AI engine and application.**
 
 
 [![Website][website-shield]][website-url]
@@ -31,16 +31,10 @@
 [github-downloads-latest-shield]: https://img.shields.io/github/downloads/comfyanonymous/ComfyUI/latest/total?style=flat&label=downloads%40latest
 [github-downloads-link]: https://github.com/comfyanonymous/ComfyUI/releases
 
-<img width="1590" height="795" alt="ComfyUI Screenshot" src="https://github.com/user-attachments/assets/36e065e0-bfae-4456-8c7f-8369d5ea48a2" />
-<br>
+![ComfyUI Screenshot](https://github.com/user-attachments/assets/7ccaf2c1-9b72-41ae-9a89-5688c94b7abe)
 </div>
 
-ComfyUI is the AI creation engine for visual professionals who demand control over every model, every parameter, and every output. Its powerful and modular node graph interface empowers creatives to generate images, videos, 3D models, audio, and more...
-- ComfyUI natively supports the latest open-source state of the art models.
-- API nodes provide access to the best closed source models such as Nano Banana, Seedance, Hunyuan3D, etc.
-- It is available on Windows, Linux, and macOS, locally with our desktop application or on our cloud.
-- The most sophisticated workflows can be exposed through a simple UI thanks to App Mode.
-- It integrates seamlessly into production pipelines with our API endpoints.
+ComfyUI lets you design and execute advanced stable diffusion pipelines using a graph/nodes/flowchart based interface. Available on Windows, Linux, and macOS.
 
 ## Get Started
 
@@ -67,7 +61,6 @@ See what ComfyUI can do with the [newer template workflows](https://comfy.org/wo
 
 ## Features
 - Nodes/graph/flowchart interface to experiment and create complex Stable Diffusion workflows without needing to code anything.
-- NOTE: There are many more models supported than the list below, if you want to see what is supported see our templates list inside ComfyUI.
 - Image Models
    - SD1.x, SD2.x ([unCLIP](https://comfyanonymous.github.io/ComfyUI_examples/unclip/))
    - [SDXL](https://comfyanonymous.github.io/ComfyUI_examples/sdxl/), [SDXL Turbo](https://comfyanonymous.github.io/ComfyUI_examples/sdturbo/)
@@ -83,7 +76,6 @@ See what ComfyUI can do with the [newer template workflows](https://comfy.org/wo
    - [Hunyuan Image 2.1](https://comfyanonymous.github.io/ComfyUI_examples/hunyuan_image/)
    - [Flux 2](https://comfyanonymous.github.io/ComfyUI_examples/flux2/)
    - [Z Image](https://comfyanonymous.github.io/ComfyUI_examples/z_image/)
-   - Ernie Image
 - Image Editing Models
    - [Omnigen 2](https://comfyanonymous.github.io/ComfyUI_examples/omnigen/)
    - [Flux Kontext](https://comfyanonymous.github.io/ComfyUI_examples/flux/#flux-kontext-image-editing-model)
@@ -133,7 +125,7 @@ Workflow examples can be found on the [Examples page](https://comfyanonymous.git
 ComfyUI follows a weekly release cycle targeting Monday but this regularly changes because of model releases or large changes to the codebase. There are three interconnected repositories:
 
 1. **[ComfyUI Core](https://github.com/comfyanonymous/ComfyUI)**
-   - Releases a new major stable version (e.g., v0.7.0) roughly every 2 weeks.
+   - Releases a new stable version (e.g., v0.7.0) roughly every week.
    - Starting from v0.4.0 patch versions will be used for fixes backported onto the current stable release.
    - Minor versions will be used for releases off the master branch.
    - Patch versions may still be used for releases on the master branch in cases where a backport would not make sense.
@@ -144,7 +136,7 @@ ComfyUI follows a weekly release cycle targeting Monday but this regularly chang
    - Builds a new release using the latest stable core version
 
 3. **[ComfyUI Frontend](https://github.com/Comfy-Org/ComfyUI_frontend)**
-   - Every 2+ weeks frontend updates are merged into the core repository
+   - Weekly frontend updates are merged into the core repository
    - Features are frozen for the upcoming core release
    - Development continues for the next release cycle
 
@@ -200,15 +192,11 @@ If you have trouble extracting it, right click the file -> properties -> unblock
 
 The portable above currently comes with python 3.13 and pytorch cuda 13.0. Update your Nvidia drivers if it doesn't start.
 
-#### All Official Portable Downloads:
+#### Alternative Downloads:
 
-[Portable for AMD GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_amd.7z)
+[Experimental portable for AMD GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_amd.7z)
 
-[Portable for Intel GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_intel.7z)
-
-[Portable for Nvidia GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia.7z) (supports 20 series and above).
-
-[Portable for Nvidia GPUs with pytorch cuda 12.6 and python 3.12](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia_cu126.7z) (Supports Nvidia 10 series and older GPUs).
+[Portable with pytorch cuda 12.6 and python 3.12](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia_cu126.7z) (Supports Nvidia 10 series and older GPUs).
 
 #### How do I share models between another UI and ComfyUI?
 
@@ -287,7 +275,7 @@ Nvidia users should install stable pytorch using this command:
 
 This is the command to install pytorch nightly instead which might have performance improvements.
 
-```pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu132```
+```pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu130```
 
 #### Troubleshooting
 

api_server/routes/internal/internal_routes.py

@@ -67,7 +67,7 @@ class InternalRoutes:
                 (entry for entry in os.scandir(directory) if is_visible_file(entry)),
                 key=lambda entry: -entry.stat().st_mtime
             )
-            return web.json_response([f"{entry.name} [{directory_type}]" for entry in sorted_files], status=200)
+            return web.json_response([entry.name for entry in sorted_files], status=200)
 
 
     def get_app(self):

app/frontend_management.py

@@ -27,7 +27,7 @@ def frontend_install_warning_message():
     return f"""
 {get_missing_requirements_message()}
 
-The ComfyUI frontend is shipped in a pip package so it needs to be updated separately from the ComfyUI code.
+This error is happening because the ComfyUI frontend is no longer shipped as part of the main repo but as a pip package instead.
 """.strip()
 
 def parse_version(version: str) -> tuple[int, int, int]:

app/node_replace_manager.py

@@ -1,7 +1,5 @@
 from __future__ import annotations
 
-import logging
-
 from aiohttp import web
 
 from typing import TYPE_CHECKING, TypedDict
@@ -33,22 +31,8 @@ class NodeReplaceManager:
         self._replacements: dict[str, list[NodeReplace]] = {}
 
     def register(self, node_replace: NodeReplace):
-        """Register a node replacement mapping.
-
-        Idempotent: if a replacement with the same (old_node_id, new_node_id)
-        is already registered, the duplicate is ignored. This prevents stale
-        entries from accumulating when custom nodes are reloaded in the same
-        process (e.g. via ComfyUI-Manager).
-        """
-        existing = self._replacements.setdefault(node_replace.old_node_id, [])
-        for entry in existing:
-            if entry.new_node_id == node_replace.new_node_id:
-                logging.debug(
-                    "Node replacement %s -> %s already registered, ignoring duplicate.",
-                    node_replace.old_node_id, node_replace.new_node_id,
-                )
-                return
-        existing.append(node_replace)
+        """Register a node replacement mapping."""
+        self._replacements.setdefault(node_replace.old_node_id, []).append(node_replace)
 
     def get_replacement(self, old_node_id: str) -> list[NodeReplace] | None:
         """Get replacements for an old node ID."""

app/user_manager.py

@@ -28,8 +28,8 @@ def get_file_info(path: str, relative_to: str) -> FileInfo:
     return {
         "path": os.path.relpath(path, relative_to).replace(os.sep, '/'),
         "size": os.path.getsize(path),
-        "modified": int(os.path.getmtime(path) * 1000),
-        "created": int(os.path.getctime(path) * 1000),
+        "modified": os.path.getmtime(path),
+        "created": os.path.getctime(path)
     }
 
 

blueprints/.glsl/Glow_30.frag

@@ -2,6 +2,7 @@
 precision mediump float;
 
 uniform sampler2D u_image0;
+uniform vec2 u_resolution;
 uniform int u_int0;      // Blend mode
 uniform int u_int1;      // Color tint
 uniform float u_float0;  // Intensity
@@ -74,7 +75,7 @@ void main() {
     float t0 = threshold - 0.15;
     float t1 = threshold + 0.15;
     
-    vec2 texelSize = 1.0 / vec2(textureSize(u_image0, 0));
+    vec2 texelSize = 1.0 / u_resolution;
     float radius2 = radius * radius;
     
     float sampleScale = clamp(radius * 0.75, 0.35, 1.0);

blueprints/.glsl/Image_Blur_1.frag

@@ -12,6 +12,7 @@ const int RADIAL_SAMPLES = 12;
 const float RADIAL_STRENGTH = 0.0003;
 
 uniform sampler2D u_image0;
+uniform vec2 u_resolution;
 uniform int u_int0;      // Blur type (BLUR_GAUSSIAN, BLUR_BOX, BLUR_RADIAL)
 uniform float u_float0;  // Blur radius/amount
 uniform int u_pass;      // Pass index (0 = horizontal, 1 = vertical)
@@ -24,7 +25,7 @@ float gaussian(float x, float sigma) {
 }
 
 void main() {
-    vec2 texelSize = 1.0 / vec2(textureSize(u_image0, 0));
+    vec2 texelSize = 1.0 / u_resolution;
     float radius = max(u_float0, 0.0);
 
     // Radial (angular) blur - single pass, doesn't use separable

blueprints/.glsl/Sharpen_23.frag

@@ -2,13 +2,14 @@
 precision highp float;
 
 uniform sampler2D u_image0;
+uniform vec2 u_resolution;
 uniform float u_float0;  // strength [0.0 – 2.0] typical: 0.3–1.0
 
 in vec2 v_texCoord;
 layout(location = 0) out vec4 fragColor0;
 
 void main() {
-    vec2 texel = 1.0 / vec2(textureSize(u_image0, 0));
+    vec2 texel = 1.0 / u_resolution;
     
     // Sample center and neighbors
     vec4 center = texture(u_image0, v_texCoord);

blueprints/.glsl/Unsharp_Mask_26.frag

@@ -2,6 +2,7 @@
 precision highp float;
 
 uniform sampler2D u_image0;
+uniform vec2 u_resolution;
 uniform float u_float0;  // amount    [0.0 - 3.0]  typical: 0.5-1.5
 uniform float u_float1;  // radius    [0.5 - 10.0] blur radius in pixels
 uniform float u_float2;  // threshold [0.0 - 0.1]  min difference to sharpen
@@ -18,7 +19,7 @@ float getLuminance(vec3 color) {
 }
 
 void main() {
-    vec2 texel = 1.0 / vec2(textureSize(u_image0, 0));
+    vec2 texel = 1.0 / u_resolution;
     float radius = max(u_float1, 0.5);
     float amount = u_float0;
     float threshold = u_float2;

blueprints/Frame Interpolation.json

@@ -1,858 +0,0 @@
-{
-  "revision": 0,
-  "last_node_id": 16,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 16,
-      "type": "022693be-2baa-4009-870a-28921508a7ef",
-      "pos": [
-        -2990,
-        -3240
-      ],
-      "size": [
-        410,
-        200
-      ],
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "inputs": [
-        {
-          "localized_name": "video",
-          "name": "video",
-          "type": "VIDEO",
-          "link": null
-        },
-        {
-          "label": "multiplier",
-          "name": "value",
-          "type": "INT",
-          "widget": {
-            "name": "value"
-          },
-          "link": null
-        },
-        {
-          "label": "enable_fps_multiplier",
-          "name": "value_1",
-          "type": "BOOLEAN",
-          "widget": {
-            "name": "value_1"
-          },
-          "link": null
-        },
-        {
-          "name": "model_name",
-          "type": "COMBO",
-          "widget": {
-            "name": "model_name"
-          },
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "label": "VIDEO",
-          "name": "VIDEO_1",
-          "type": "VIDEO",
-          "links": []
-        },
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": null
-        }
-      ],
-      "properties": {
-        "proxyWidgets": [
-          [
-            "9",
-            "value"
-          ],
-          [
-            "13",
-            "value"
-          ],
-          [
-            "1",
-            "model_name"
-          ]
-        ],
-        "enableTabs": false,
-        "tabWidth": 65,
-        "tabXOffset": 10,
-        "hasSecondTab": false,
-        "secondTabText": "Send Back",
-        "secondTabOffset": 80,
-        "secondTabWidth": 65,
-        "cnr_id": "comfy-core",
-        "ver": "0.19.3"
-      },
-      "widgets_values": [],
-      "title": "Frame Interpolation"
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "022693be-2baa-4009-870a-28921508a7ef",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 17,
-          "lastLinkId": 28,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Frame Interpolation",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            -2810,
-            -3070,
-            159.7421875,
-            120
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            -1270,
-            -3075,
-            120,
-            80
-          ]
-        },
-        "inputs": [
-          {
-            "id": "05e31c51-dcb6-4a1e-9651-1b9ad4f7a287",
-            "name": "video",
-            "type": "VIDEO",
-            "linkIds": [
-              2
-            ],
-            "localized_name": "video",
-            "pos": [
-              -2670.2578125,
-              -3050
-            ]
-          },
-          {
-            "id": "feecb409-7d1c-4a99-9c63-50c5fecdd3c9",
-            "name": "value",
-            "type": "INT",
-            "linkIds": [
-              22
-            ],
-            "label": "multiplier",
-            "pos": [
-              -2670.2578125,
-              -3030
-            ]
-          },
-          {
-            "id": "0b8a861b-b581-4068-9e8c-f8d15daf1ca6",
-            "name": "value_1",
-            "type": "BOOLEAN",
-            "linkIds": [
-              23
-            ],
-            "label": "enable_fps_multiplier",
-            "pos": [
-              -2670.2578125,
-              -3010
-            ]
-          },
-          {
-            "id": "a22b101e-8773-4e17-a297-7ee3aae09162",
-            "name": "model_name",
-            "type": "COMBO",
-            "linkIds": [
-              24
-            ],
-            "pos": [
-              -2670.2578125,
-              -2990
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "ef2ada05-d5aa-492a-9394-6c3e71e39ebb",
-            "name": "VIDEO_1",
-            "type": "VIDEO",
-            "linkIds": [
-              26
-            ],
-            "label": "VIDEO",
-            "pos": [
-              -1250,
-              -3055
-            ]
-          },
-          {
-            "id": "5aacc622-2a07-4983-b31c-e04461f7f953",
-            "name": "IMAGE",
-            "type": "IMAGE",
-            "linkIds": [
-              28
-            ],
-            "pos": [
-              -1250,
-              -3035
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 1,
-            "type": "FrameInterpolationModelLoader",
-            "pos": [
-              -2510,
-              -3370
-            ],
-            "size": [
-              370,
-              90
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "model_name",
-                "name": "model_name",
-                "type": "COMBO",
-                "widget": {
-                  "name": "model_name"
-                },
-                "link": 24
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "INTERP_MODEL",
-                "name": "INTERP_MODEL",
-                "type": "INTERP_MODEL",
-                "links": [
-                  1
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "FrameInterpolationModelLoader",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "models": [
-                {
-                  "name": "film_net_fp16.safetensors",
-                  "url": "https://huggingface.co/Comfy-Org/frame_interpolation/resolve/main/frame_interpolation/film_net_fp16.safetensors",
-                  "directory": "frame_interpolation"
-                }
-              ]
-            },
-            "widgets_values": [
-              "film_net_fp16.safetensors"
-            ]
-          },
-          {
-            "id": 2,
-            "type": "FrameInterpolate",
-            "pos": [
-              -2040,
-              -3370
-            ],
-            "size": [
-              270,
-              110
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "interp_model",
-                "name": "interp_model",
-                "type": "INTERP_MODEL",
-                "link": 1
-              },
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "link": 3
-              },
-              {
-                "localized_name": "multiplier",
-                "name": "multiplier",
-                "type": "INT",
-                "widget": {
-                  "name": "multiplier"
-                },
-                "link": 8
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "IMAGE",
-                "name": "IMAGE",
-                "type": "IMAGE",
-                "links": [
-                  4,
-                  28
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "FrameInterpolate",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            },
-            "widgets_values": [
-              2
-            ]
-          },
-          {
-            "id": 5,
-            "type": "CreateVideo",
-            "pos": [
-              -1600,
-              -3370
-            ],
-            "size": [
-              270,
-              110
-            ],
-            "flags": {},
-            "order": 3,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "link": 4
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "shape": 7,
-                "type": "AUDIO",
-                "link": 5
-              },
-              {
-                "localized_name": "fps",
-                "name": "fps",
-                "type": "FLOAT",
-                "widget": {
-                  "name": "fps"
-                },
-                "link": 12
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "VIDEO",
-                "name": "VIDEO",
-                "type": "VIDEO",
-                "links": [
-                  26
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "CreateVideo",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            },
-            "widgets_values": [
-              30
-            ]
-          },
-          {
-            "id": 9,
-            "type": "PrimitiveInt",
-            "pos": [
-              -2500,
-              -2970
-            ],
-            "size": [
-              270,
-              90
-            ],
-            "flags": {},
-            "order": 4,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "value",
-                "name": "value",
-                "type": "INT",
-                "widget": {
-                  "name": "value"
-                },
-                "link": 22
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "INT",
-                "name": "INT",
-                "type": "INT",
-                "links": [
-                  8,
-                  19
-                ]
-              }
-            ],
-            "title": "Int (Multiplier)",
-            "properties": {
-              "Node name for S&R": "PrimitiveInt",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            },
-            "widgets_values": [
-              2,
-              "fixed"
-            ]
-          },
-          {
-            "id": 10,
-            "type": "ComfySwitchNode",
-            "pos": [
-              -1610,
-              -3120
-            ],
-            "size": [
-              270,
-              130
-            ],
-            "flags": {},
-            "order": 5,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "on_false",
-                "name": "on_false",
-                "type": "*",
-                "link": 11
-              },
-              {
-                "localized_name": "on_true",
-                "name": "on_true",
-                "type": "*",
-                "link": 13
-              },
-              {
-                "localized_name": "switch",
-                "name": "switch",
-                "type": "BOOLEAN",
-                "widget": {
-                  "name": "switch"
-                },
-                "link": 15
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "output",
-                "name": "output",
-                "type": "*",
-                "links": [
-                  12
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "ComfySwitchNode",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            },
-            "widgets_values": [
-              true
-            ]
-          },
-          {
-            "id": 13,
-            "type": "PrimitiveBoolean",
-            "pos": [
-              -2500,
-              -2770
-            ],
-            "size": [
-              310,
-              90
-            ],
-            "flags": {},
-            "order": 7,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "value",
-                "name": "value",
-                "type": "BOOLEAN",
-                "widget": {
-                  "name": "value"
-                },
-                "link": 23
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "BOOLEAN",
-                "name": "BOOLEAN",
-                "type": "BOOLEAN",
-                "links": [
-                  15
-                ]
-              }
-            ],
-            "title": "Boolean (Apply multiplier to FPS?)",
-            "properties": {
-              "Node name for S&R": "PrimitiveBoolean",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            },
-            "widgets_values": [
-              true
-            ]
-          },
-          {
-            "id": 3,
-            "type": "GetVideoComponents",
-            "pos": [
-              -2500,
-              -3170
-            ],
-            "size": [
-              230,
-              100
-            ],
-            "flags": {},
-            "order": 2,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "video",
-                "name": "video",
-                "type": "VIDEO",
-                "link": 2
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "links": [
-                  3
-                ]
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "type": "AUDIO",
-                "links": [
-                  5
-                ]
-              },
-              {
-                "localized_name": "fps",
-                "name": "fps",
-                "type": "FLOAT",
-                "links": [
-                  11,
-                  18
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "GetVideoComponents",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            }
-          },
-          {
-            "id": 11,
-            "type": "ComfyMathExpression",
-            "pos": [
-              -2090,
-              -3070
-            ],
-            "size": [
-              400,
-              210
-            ],
-            "flags": {
-              "collapsed": false
-            },
-            "order": 6,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "a",
-                "localized_name": "values.a",
-                "name": "values.a",
-                "type": "FLOAT,INT",
-                "link": 18
-              },
-              {
-                "label": "b",
-                "localized_name": "values.b",
-                "name": "values.b",
-                "shape": 7,
-                "type": "FLOAT,INT",
-                "link": 19
-              },
-              {
-                "label": "c",
-                "localized_name": "values.c",
-                "name": "values.c",
-                "shape": 7,
-                "type": "FLOAT,INT",
-                "link": null
-              },
-              {
-                "localized_name": "expression",
-                "name": "expression",
-                "type": "STRING",
-                "widget": {
-                  "name": "expression"
-                },
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "FLOAT",
-                "name": "FLOAT",
-                "type": "FLOAT",
-                "links": [
-                  13
-                ]
-              },
-              {
-                "localized_name": "INT",
-                "name": "INT",
-                "type": "INT",
-                "links": null
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "ComfyMathExpression",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3"
-            },
-            "widgets_values": [
-              "min(abs(b), 16) * a"
-            ]
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 1,
-            "origin_id": 1,
-            "origin_slot": 0,
-            "target_id": 2,
-            "target_slot": 0,
-            "type": "INTERP_MODEL"
-          },
-          {
-            "id": 3,
-            "origin_id": 3,
-            "origin_slot": 0,
-            "target_id": 2,
-            "target_slot": 1,
-            "type": "IMAGE"
-          },
-          {
-            "id": 8,
-            "origin_id": 9,
-            "origin_slot": 0,
-            "target_id": 2,
-            "target_slot": 2,
-            "type": "INT"
-          },
-          {
-            "id": 4,
-            "origin_id": 2,
-            "origin_slot": 0,
-            "target_id": 5,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 5,
-            "origin_id": 3,
-            "origin_slot": 1,
-            "target_id": 5,
-            "target_slot": 1,
-            "type": "AUDIO"
-          },
-          {
-            "id": 12,
-            "origin_id": 10,
-            "origin_slot": 0,
-            "target_id": 5,
-            "target_slot": 2,
-            "type": "FLOAT"
-          },
-          {
-            "id": 11,
-            "origin_id": 3,
-            "origin_slot": 2,
-            "target_id": 10,
-            "target_slot": 0,
-            "type": "FLOAT"
-          },
-          {
-            "id": 13,
-            "origin_id": 11,
-            "origin_slot": 0,
-            "target_id": 10,
-            "target_slot": 1,
-            "type": "FLOAT"
-          },
-          {
-            "id": 15,
-            "origin_id": 13,
-            "origin_slot": 0,
-            "target_id": 10,
-            "target_slot": 2,
-            "type": "BOOLEAN"
-          },
-          {
-            "id": 18,
-            "origin_id": 3,
-            "origin_slot": 2,
-            "target_id": 11,
-            "target_slot": 0,
-            "type": "FLOAT"
-          },
-          {
-            "id": 19,
-            "origin_id": 9,
-            "origin_slot": 0,
-            "target_id": 11,
-            "target_slot": 1,
-            "type": "INT"
-          },
-          {
-            "id": 2,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 3,
-            "target_slot": 0,
-            "type": "VIDEO"
-          },
-          {
-            "id": 22,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 9,
-            "target_slot": 0,
-            "type": "INT"
-          },
-          {
-            "id": 23,
-            "origin_id": -10,
-            "origin_slot": 2,
-            "target_id": 13,
-            "target_slot": 0,
-            "type": "BOOLEAN"
-          },
-          {
-            "id": 24,
-            "origin_id": -10,
-            "origin_slot": 3,
-            "target_id": 1,
-            "target_slot": 0,
-            "type": "COMBO"
-          },
-          {
-            "id": 26,
-            "origin_id": 5,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "VIDEO"
-          },
-          {
-            "id": 28,
-            "origin_id": 2,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 1,
-            "type": "IMAGE"
-          }
-        ],
-        "extra": {},
-        "category": "Video Tools",
-        "description": "Increases video frame rate by synthesizing intermediate frames with a frame interpolation model."
-      }
-    ]
-  },
-  "extra": {}
-}

blueprints/Get Any Video Frame.json

@@ -1,485 +0,0 @@
-{
-  "revision": 0,
-  "last_node_id": 98,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 98,
-      "type": "dca6e78d-fb06-421e-97f7-6ce17a665260",
-      "pos": [
-        -410,
-        -2230
-      ],
-      "size": [
-        270,
-        104
-      ],
-      "flags": {},
-      "order": 7,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "video",
-          "type": "VIDEO",
-          "link": null
-        },
-        {
-          "label": "frame_index",
-          "name": "value",
-          "type": "INT",
-          "widget": {
-            "name": "value"
-          },
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": []
-        }
-      ],
-      "title": "Get Any Video Frame",
-      "properties": {
-        "proxyWidgets": [
-          [
-            "100",
-            "value"
-          ]
-        ]
-      },
-      "widgets_values": []
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "dca6e78d-fb06-421e-97f7-6ce17a665260",
-        "version": 1,
-        "state": {
-          "lastGroupId": 1,
-          "lastNodeId": 136,
-          "lastLinkId": 302,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Get Any Video Frame",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            380,
-            -57,
-            120,
-            80
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            1460,
-            -57,
-            120,
-            60
-          ]
-        },
-        "inputs": [
-          {
-            "id": "2ceec378-8dcf-4340-8570-155967f59a93",
-            "name": "video",
-            "type": "VIDEO",
-            "linkIds": [
-              4
-            ],
-            "pos": [
-              480,
-              -37
-            ]
-          },
-          {
-            "id": "819955f6-c686-4896-8032-ff2d0059109a",
-            "name": "value",
-            "type": "INT",
-            "linkIds": [
-              283
-            ],
-            "label": "frame_index",
-            "pos": [
-              480,
-              -17
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "1ab0684d-6a44-45b6-8aa4-a0b971a1d41e",
-            "name": "IMAGE",
-            "type": "IMAGE",
-            "linkIds": [
-              5
-            ],
-            "pos": [
-              1480,
-              -37
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 1,
-            "type": "GetVideoComponents",
-            "pos": [
-              560,
-              -150
-            ],
-            "size": [
-              230,
-              120
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "video",
-                "name": "video",
-                "type": "VIDEO",
-                "link": 4
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "links": [
-                  1,
-                  2
-                ]
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "type": "AUDIO",
-                "links": null
-              },
-              {
-                "localized_name": "fps",
-                "name": "fps",
-                "type": "FLOAT",
-                "links": null
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "GetVideoComponents"
-            }
-          },
-          {
-            "id": 2,
-            "type": "GetImageSize",
-            "pos": [
-              560,
-              50
-            ],
-            "size": [
-              230,
-              120
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 1
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "width",
-                "name": "width",
-                "type": "INT",
-                "links": null
-              },
-              {
-                "localized_name": "height",
-                "name": "height",
-                "type": "INT",
-                "links": null
-              },
-              {
-                "localized_name": "batch_size",
-                "name": "batch_size",
-                "type": "INT",
-                "links": [
-                  285
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "GetImageSize"
-            }
-          },
-          {
-            "id": 3,
-            "type": "ImageFromBatch",
-            "pos": [
-              1130,
-              -150
-            ],
-            "size": [
-              270,
-              140
-            ],
-            "flags": {},
-            "order": 2,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 2
-              },
-              {
-                "localized_name": "batch_index",
-                "name": "batch_index",
-                "type": "INT",
-                "widget": {
-                  "name": "batch_index"
-                },
-                "link": 286
-              },
-              {
-                "localized_name": "length",
-                "name": "length",
-                "type": "INT",
-                "widget": {
-                  "name": "length"
-                },
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "IMAGE",
-                "name": "IMAGE",
-                "type": "IMAGE",
-                "links": [
-                  5
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "ImageFromBatch"
-            },
-            "widgets_values": [
-              0,
-              1
-            ]
-          },
-          {
-            "id": 99,
-            "type": "ComfyMathExpression",
-            "pos": [
-              910,
-              100
-            ],
-            "size": [
-              400,
-              200
-            ],
-            "flags": {},
-            "order": 3,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "a",
-                "localized_name": "values.a",
-                "name": "values.a",
-                "type": "FLOAT,INT",
-                "link": 284
-              },
-              {
-                "label": "b",
-                "localized_name": "values.b",
-                "name": "values.b",
-                "shape": 7,
-                "type": "FLOAT,INT",
-                "link": 285
-              },
-              {
-                "label": "c",
-                "localized_name": "values.c",
-                "name": "values.c",
-                "shape": 7,
-                "type": "FLOAT,INT",
-                "link": null
-              },
-              {
-                "localized_name": "expression",
-                "name": "expression",
-                "type": "STRING",
-                "widget": {
-                  "name": "expression"
-                },
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "FLOAT",
-                "name": "FLOAT",
-                "type": "FLOAT",
-                "links": null
-              },
-              {
-                "localized_name": "INT",
-                "name": "INT",
-                "type": "INT",
-                "links": [
-                  286
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "ComfyMathExpression"
-            },
-            "widgets_values": [
-              "min(max(int(a if a >= 0 else b + a), 0), b - 1)"
-            ]
-          },
-          {
-            "id": 100,
-            "type": "PrimitiveInt",
-            "pos": [
-              560,
-              250
-            ],
-            "size": [
-              270,
-              110
-            ],
-            "flags": {},
-            "order": 4,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "value",
-                "name": "value",
-                "type": "INT",
-                "widget": {
-                  "name": "value"
-                },
-                "link": 283
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "INT",
-                "name": "INT",
-                "type": "INT",
-                "links": [
-                  284
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "PrimitiveInt"
-            },
-            "widgets_values": [
-              0,
-              "fixed"
-            ]
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 1,
-            "origin_id": 1,
-            "origin_slot": 0,
-            "target_id": 2,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 2,
-            "origin_id": 1,
-            "origin_slot": 0,
-            "target_id": 3,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 4,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 1,
-            "target_slot": 0,
-            "type": "VIDEO"
-          },
-          {
-            "id": 5,
-            "origin_id": 3,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 283,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 100,
-            "target_slot": 0,
-            "type": "INT"
-          },
-          {
-            "id": 284,
-            "origin_id": 100,
-            "origin_slot": 0,
-            "target_id": 99,
-            "target_slot": 0,
-            "type": "INT"
-          },
-          {
-            "id": 285,
-            "origin_id": 2,
-            "origin_slot": 2,
-            "target_id": 99,
-            "target_slot": 1,
-            "type": "INT"
-          },
-          {
-            "id": 286,
-            "origin_id": 99,
-            "origin_slot": 1,
-            "target_id": 3,
-            "target_slot": 1,
-            "type": "INT"
-          }
-        ],
-        "extra": {},
-        "category": "Video Tools",
-        "description": "Extracts one image frame from a video at a chosen index, with optional trim and FPS control."
-      }
-    ]
-  },
-  "extra": {
-    "ds": {
-      "scale": 1.197015527856339,
-      "offset": [
-        -168.76833554248222,
-        540.6638955283997
-      ]
-    },
-    "frontendVersion": "1.42.8"
-  }
-}

blueprints/Image Channels.json

@@ -1,323 +1 @@
-{
-  "revision": 0,
-  "last_node_id": 29,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 29,
-      "type": "4c9d6ea4-b912-40e5-8766-6793a9758c53",
-      "pos": [
-        1970,
-        -230
-      ],
-      "size": [
-        180,
-        86
-      ],
-      "flags": {},
-      "order": 5,
-      "mode": 0,
-      "inputs": [
-        {
-          "label": "image",
-          "localized_name": "images.image0",
-          "name": "images.image0",
-          "type": "IMAGE",
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "label": "R",
-          "localized_name": "IMAGE0",
-          "name": "IMAGE0",
-          "type": "IMAGE",
-          "links": []
-        },
-        {
-          "label": "G",
-          "localized_name": "IMAGE1",
-          "name": "IMAGE1",
-          "type": "IMAGE",
-          "links": []
-        },
-        {
-          "label": "B",
-          "localized_name": "IMAGE2",
-          "name": "IMAGE2",
-          "type": "IMAGE",
-          "links": []
-        },
-        {
-          "label": "A",
-          "localized_name": "IMAGE3",
-          "name": "IMAGE3",
-          "type": "IMAGE",
-          "links": []
-        }
-      ],
-      "title": "Image Channels",
-      "properties": {
-        "proxyWidgets": []
-      },
-      "widgets_values": []
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "4c9d6ea4-b912-40e5-8766-6793a9758c53",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 28,
-          "lastLinkId": 39,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Image Channels",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            1820,
-            -185,
-            120,
-            60
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            2460,
-            -215,
-            120,
-            120
-          ]
-        },
-        "inputs": [
-          {
-            "id": "3522932b-2d86-4a1f-a02a-cb29f3a9d7fe",
-            "name": "images.image0",
-            "type": "IMAGE",
-            "linkIds": [
-              39
-            ],
-            "localized_name": "images.image0",
-            "label": "image",
-            "pos": [
-              1920,
-              -165
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "605cb9c3-b065-4d9b-81d2-3ec331889b2b",
-            "name": "IMAGE0",
-            "type": "IMAGE",
-            "linkIds": [
-              26
-            ],
-            "localized_name": "IMAGE0",
-            "label": "R",
-            "pos": [
-              2480,
-              -195
-            ]
-          },
-          {
-            "id": "fb44a77e-0522-43e9-9527-82e7465b3596",
-            "name": "IMAGE1",
-            "type": "IMAGE",
-            "linkIds": [
-              27
-            ],
-            "localized_name": "IMAGE1",
-            "label": "G",
-            "pos": [
-              2480,
-              -175
-            ]
-          },
-          {
-            "id": "81460ee6-0131-402a-874f-6bf3001fc4ff",
-            "name": "IMAGE2",
-            "type": "IMAGE",
-            "linkIds": [
-              28
-            ],
-            "localized_name": "IMAGE2",
-            "label": "B",
-            "pos": [
-              2480,
-              -155
-            ]
-          },
-          {
-            "id": "ae690246-80d4-4951-b1d9-9306d8a77417",
-            "name": "IMAGE3",
-            "type": "IMAGE",
-            "linkIds": [
-              29
-            ],
-            "localized_name": "IMAGE3",
-            "label": "A",
-            "pos": [
-              2480,
-              -135
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 23,
-            "type": "GLSLShader",
-            "pos": [
-              2000,
-              -330
-            ],
-            "size": [
-              400,
-              172
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "image",
-                "localized_name": "images.image0",
-                "name": "images.image0",
-                "type": "IMAGE",
-                "link": 39
-              },
-              {
-                "localized_name": "fragment_shader",
-                "name": "fragment_shader",
-                "type": "STRING",
-                "widget": {
-                  "name": "fragment_shader"
-                },
-                "link": null
-              },
-              {
-                "localized_name": "size_mode",
-                "name": "size_mode",
-                "type": "COMFY_DYNAMICCOMBO_V3",
-                "widget": {
-                  "name": "size_mode"
-                },
-                "link": null
-              },
-              {
-                "label": "image1",
-                "localized_name": "images.image1",
-                "name": "images.image1",
-                "shape": 7,
-                "type": "IMAGE",
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "label": "R",
-                "localized_name": "IMAGE0",
-                "name": "IMAGE0",
-                "type": "IMAGE",
-                "links": [
-                  26
-                ]
-              },
-              {
-                "label": "G",
-                "localized_name": "IMAGE1",
-                "name": "IMAGE1",
-                "type": "IMAGE",
-                "links": [
-                  27
-                ]
-              },
-              {
-                "label": "B",
-                "localized_name": "IMAGE2",
-                "name": "IMAGE2",
-                "type": "IMAGE",
-                "links": [
-                  28
-                ]
-              },
-              {
-                "label": "A",
-                "localized_name": "IMAGE3",
-                "name": "IMAGE3",
-                "type": "IMAGE",
-                "links": [
-                  29
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "GLSLShader"
-            },
-            "widgets_values": [
-              "#version 300 es\nprecision highp float;\n\nuniform sampler2D u_image0;\n\nin vec2 v_texCoord;\nlayout(location = 0) out vec4 fragColor0;\nlayout(location = 1) out vec4 fragColor1;\nlayout(location = 2) out vec4 fragColor2;\nlayout(location = 3) out vec4 fragColor3;\n\nvoid main() {\n  vec4 color = texture(u_image0, v_texCoord);\n  // Output each channel as grayscale to separate render targets\n  fragColor0 = vec4(vec3(color.r), 1.0);  // Red channel\n  fragColor1 = vec4(vec3(color.g), 1.0);  // Green channel\n  fragColor2 = vec4(vec3(color.b), 1.0);  // Blue channel\n  fragColor3 = vec4(vec3(color.a), 1.0);  // Alpha channel\n}\n",
-              "from_input"
-            ]
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 39,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 23,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 26,
-            "origin_id": 23,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 27,
-            "origin_id": 23,
-            "origin_slot": 1,
-            "target_id": -20,
-            "target_slot": 1,
-            "type": "IMAGE"
-          },
-          {
-            "id": 28,
-            "origin_id": 23,
-            "origin_slot": 2,
-            "target_id": -20,
-            "target_slot": 2,
-            "type": "IMAGE"
-          },
-          {
-            "id": 29,
-            "origin_id": 23,
-            "origin_slot": 3,
-            "target_id": -20,
-            "target_slot": 3,
-            "type": "IMAGE"
-          }
-        ],
-        "extra": {
-          "workflowRendererVersion": "LG"
-        },
-        "category": "Image Tools/Color adjust",
-        "description": "Manipulates individual RGBA channels for masking, compositing, and channel effects."
-      }
-    ]
-  }
-}
+{"revision": 0, "last_node_id": 29, "last_link_id": 0, "nodes": [{"id": 29, "type": "4c9d6ea4-b912-40e5-8766-6793a9758c53", "pos": [1970, -230], "size": [180, 86], "flags": {}, "order": 5, "mode": 0, "inputs": [{"label": "image", "localized_name": "images.image0", "name": "images.image0", "type": "IMAGE", "link": null}], "outputs": [{"label": "R", "localized_name": "IMAGE0", "name": "IMAGE0", "type": "IMAGE", "links": []}, {"label": "G", "localized_name": "IMAGE1", "name": "IMAGE1", "type": "IMAGE", "links": []}, {"label": "B", "localized_name": "IMAGE2", "name": "IMAGE2", "type": "IMAGE", "links": []}, {"label": "A", "localized_name": "IMAGE3", "name": "IMAGE3", "type": "IMAGE", "links": []}], "title": "Image Channels", "properties": {"proxyWidgets": []}, "widgets_values": []}], "links": [], "version": 0.4, "definitions": {"subgraphs": [{"id": "4c9d6ea4-b912-40e5-8766-6793a9758c53", "version": 1, "state": {"lastGroupId": 0, "lastNodeId": 28, "lastLinkId": 39, "lastRerouteId": 0}, "revision": 0, "config": {}, "name": "Image Channels", "inputNode": {"id": -10, "bounding": [1820, -185, 120, 60]}, "outputNode": {"id": -20, "bounding": [2460, -215, 120, 120]}, "inputs": [{"id": "3522932b-2d86-4a1f-a02a-cb29f3a9d7fe", "name": "images.image0", "type": "IMAGE", "linkIds": [39], "localized_name": "images.image0", "label": "image", "pos": [1920, -165]}], "outputs": [{"id": "605cb9c3-b065-4d9b-81d2-3ec331889b2b", "name": "IMAGE0", "type": "IMAGE", "linkIds": [26], "localized_name": "IMAGE0", "label": "R", "pos": [2480, -195]}, {"id": "fb44a77e-0522-43e9-9527-82e7465b3596", "name": "IMAGE1", "type": "IMAGE", "linkIds": [27], "localized_name": "IMAGE1", "label": "G", "pos": [2480, -175]}, {"id": "81460ee6-0131-402a-874f-6bf3001fc4ff", "name": "IMAGE2", "type": "IMAGE", "linkIds": [28], "localized_name": "IMAGE2", "label": "B", "pos": [2480, -155]}, {"id": "ae690246-80d4-4951-b1d9-9306d8a77417", "name": "IMAGE3", "type": "IMAGE", "linkIds": [29], "localized_name": "IMAGE3", "label": "A", "pos": [2480, -135]}], "widgets": [], "nodes": [{"id": 23, "type": "GLSLShader", "pos": [2000, -330], "size": [400, 172], "flags": {}, "order": 0, "mode": 0, "inputs": [{"label": "image", "localized_name": "images.image0", "name": "images.image0", "type": "IMAGE", "link": 39}, {"localized_name": "fragment_shader", "name": "fragment_shader", "type": "STRING", "widget": {"name": "fragment_shader"}, "link": null}, {"localized_name": "size_mode", "name": "size_mode", "type": "COMFY_DYNAMICCOMBO_V3", "widget": {"name": "size_mode"}, "link": null}, {"label": "image1", "localized_name": "images.image1", "name": "images.image1", "shape": 7, "type": "IMAGE", "link": null}], "outputs": [{"label": "R", "localized_name": "IMAGE0", "name": "IMAGE0", "type": "IMAGE", "links": [26]}, {"label": "G", "localized_name": "IMAGE1", "name": "IMAGE1", "type": "IMAGE", "links": [27]}, {"label": "B", "localized_name": "IMAGE2", "name": "IMAGE2", "type": "IMAGE", "links": [28]}, {"label": "A", "localized_name": "IMAGE3", "name": "IMAGE3", "type": "IMAGE", "links": [29]}], "properties": {"Node name for S&R": "GLSLShader"}, "widgets_values": ["#version 300 es\nprecision highp float;\n\nuniform sampler2D u_image0;\n\nin vec2 v_texCoord;\nlayout(location = 0) out vec4 fragColor0;\nlayout(location = 1) out vec4 fragColor1;\nlayout(location = 2) out vec4 fragColor2;\nlayout(location = 3) out vec4 fragColor3;\n\nvoid main() {\n  vec4 color = texture(u_image0, v_texCoord);\n  // Output each channel as grayscale to separate render targets\n  fragColor0 = vec4(vec3(color.r), 1.0);  // Red channel\n  fragColor1 = vec4(vec3(color.g), 1.0);  // Green channel\n  fragColor2 = vec4(vec3(color.b), 1.0);  // Blue channel\n  fragColor3 = vec4(vec3(color.a), 1.0);  // Alpha channel\n}\n", "from_input"]}], "groups": [], "links": [{"id": 39, "origin_id": -10, "origin_slot": 0, "target_id": 23, "target_slot": 0, "type": "IMAGE"}, {"id": 26, "origin_id": 23, "origin_slot": 0, "target_id": -20, "target_slot": 0, "type": "IMAGE"}, {"id": 27, "origin_id": 23, "origin_slot": 1, "target_id": -20, "target_slot": 1, "type": "IMAGE"}, {"id": 28, "origin_id": 23, "origin_slot": 2, "target_id": -20, "target_slot": 2, "type": "IMAGE"}, {"id": 29, "origin_id": 23, "origin_slot": 3, "target_id": -20, "target_slot": 3, "type": "IMAGE"}], "extra": {"workflowRendererVersion": "LG"}, "category": "Image Tools/Color adjust"}]}}

blueprints/Image Segmentation (SAM3).json

@@ -1,714 +0,0 @@
-{
-  "revision": 0,
-  "last_node_id": 99,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 99,
-      "type": "6e7ab3ea-96aa-470f-9b94-3d9d0e01f481",
-      "pos": [
-        -1630,
-        -3270
-      ],
-      "size": [
-        290,
-        370
-      ],
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [
-        {
-          "label": "image",
-          "localized_name": "image",
-          "name": "image",
-          "type": "IMAGE",
-          "link": null
-        },
-        {
-          "label": "object",
-          "name": "text",
-          "type": "STRING",
-          "widget": {
-            "name": "text"
-          },
-          "link": null
-        },
-        {
-          "name": "bboxes",
-          "type": "BOUNDING_BOX",
-          "link": null
-        },
-        {
-          "name": "positive_coords",
-          "type": "STRING",
-          "link": null
-        },
-        {
-          "name": "negative_coords",
-          "type": "STRING",
-          "link": null
-        },
-        {
-          "name": "threshold",
-          "type": "FLOAT",
-          "widget": {
-            "name": "threshold"
-          },
-          "link": null
-        },
-        {
-          "name": "refine_iterations",
-          "type": "INT",
-          "widget": {
-            "name": "refine_iterations"
-          },
-          "link": null
-        },
-        {
-          "name": "individual_masks",
-          "type": "BOOLEAN",
-          "widget": {
-            "name": "individual_masks"
-          },
-          "link": null
-        },
-        {
-          "name": "ckpt_name",
-          "type": "COMBO",
-          "widget": {
-            "name": "ckpt_name"
-          },
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "localized_name": "masks",
-          "name": "masks",
-          "type": "MASK",
-          "links": []
-        },
-        {
-          "localized_name": "bboxes",
-          "name": "bboxes",
-          "type": "BOUNDING_BOX",
-          "links": []
-        }
-      ],
-      "properties": {
-        "proxyWidgets": [
-          [
-            "78",
-            "text"
-          ],
-          [
-            "75",
-            "threshold"
-          ],
-          [
-            "75",
-            "refine_iterations"
-          ],
-          [
-            "75",
-            "individual_masks"
-          ],
-          [
-            "77",
-            "ckpt_name"
-          ]
-        ],
-        "ue_properties": {
-          "widget_ue_connectable": {
-            "text": true
-          },
-          "version": "7.7",
-          "input_ue_unconnectable": {}
-        },
-        "cnr_id": "comfy-core",
-        "ver": "0.19.3",
-        "enableTabs": false,
-        "tabWidth": 65,
-        "tabXOffset": 10,
-        "hasSecondTab": false,
-        "secondTabText": "Send Back",
-        "secondTabOffset": 80,
-        "secondTabWidth": 65
-      },
-      "widgets_values": [],
-      "title": "Image Segmentation (SAM3)"
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "6e7ab3ea-96aa-470f-9b94-3d9d0e01f481",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 113,
-          "lastLinkId": 283,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Image Segmentation (SAM3)",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            -2260,
-            -3450,
-            136.369140625,
-            220
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            -1130,
-            -3305,
-            120,
-            80
-          ]
-        },
-        "inputs": [
-          {
-            "id": "a6e75fa2-162a-4af0-a2fd-1e9c899a5ab6",
-            "name": "image",
-            "type": "IMAGE",
-            "linkIds": [
-              264
-            ],
-            "localized_name": "image",
-            "label": "image",
-            "pos": [
-              -2143.630859375,
-              -3430
-            ]
-          },
-          {
-            "id": "3cefd304-7631-4ff6-a5a0-5a0ffb120745",
-            "name": "text",
-            "type": "STRING",
-            "linkIds": [
-              265
-            ],
-            "label": "object",
-            "pos": [
-              -2143.630859375,
-              -3410
-            ]
-          },
-          {
-            "id": "1aec91c5-d8d2-441c-928c-49c14e7e80ed",
-            "name": "bboxes",
-            "type": "BOUNDING_BOX",
-            "linkIds": [
-              266
-            ],
-            "pos": [
-              -2143.630859375,
-              -3390
-            ]
-          },
-          {
-            "id": "1ec7ce1a-8257-4719-8a81-60ebc8a98899",
-            "name": "positive_coords",
-            "type": "STRING",
-            "linkIds": [
-              267
-            ],
-            "pos": [
-              -2143.630859375,
-              -3370
-            ]
-          },
-          {
-            "id": "c65f8b87-9bd7-48be-9fc2-823431e95019",
-            "name": "negative_coords",
-            "type": "STRING",
-            "linkIds": [
-              268
-            ],
-            "pos": [
-              -2143.630859375,
-              -3350
-            ]
-          },
-          {
-            "id": "bb4ba35a-ccfe-4c37-98e5-d9b0d69585fb",
-            "name": "threshold",
-            "type": "FLOAT",
-            "linkIds": [
-              269
-            ],
-            "pos": [
-              -2143.630859375,
-              -3330
-            ]
-          },
-          {
-            "id": "b1439668-b050-490b-a5dc-fc4052c55666",
-            "name": "refine_iterations",
-            "type": "INT",
-            "linkIds": [
-              270
-            ],
-            "pos": [
-              -2143.630859375,
-              -3310
-            ]
-          },
-          {
-            "id": "86e239e5-c098-4302-b54d-d42a38bc0f89",
-            "name": "individual_masks",
-            "type": "BOOLEAN",
-            "linkIds": [
-              271
-            ],
-            "pos": [
-              -2143.630859375,
-              -3290
-            ]
-          },
-          {
-            "id": "f9e0b9d4-b2f1-4907-a4a5-305656576706",
-            "name": "ckpt_name",
-            "type": "COMBO",
-            "linkIds": [
-              272
-            ],
-            "pos": [
-              -2143.630859375,
-              -3270
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "ff50da09-1e59-4a58-9b7f-be1a00aa5913",
-            "name": "masks",
-            "type": "MASK",
-            "linkIds": [
-              231
-            ],
-            "localized_name": "masks",
-            "pos": [
-              -1110,
-              -3285
-            ]
-          },
-          {
-            "id": "8f622e40-8528-4078-b7d3-147e9f872194",
-            "name": "bboxes",
-            "type": "BOUNDING_BOX",
-            "linkIds": [
-              232
-            ],
-            "localized_name": "bboxes",
-            "pos": [
-              -1110,
-              -3265
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 75,
-            "type": "SAM3_Detect",
-            "pos": [
-              -1470,
-              -3460
-            ],
-            "size": [
-              270,
-              260
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "model",
-                "localized_name": "model",
-                "name": "model",
-                "type": "MODEL",
-                "link": 237
-              },
-              {
-                "label": "image",
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 264
-              },
-              {
-                "label": "conditioning",
-                "localized_name": "conditioning",
-                "name": "conditioning",
-                "shape": 7,
-                "type": "CONDITIONING",
-                "link": 200
-              },
-              {
-                "label": "bboxes",
-                "localized_name": "bboxes",
-                "name": "bboxes",
-                "shape": 7,
-                "type": "BOUNDING_BOX",
-                "link": 266
-              },
-              {
-                "label": "positive_coords",
-                "localized_name": "positive_coords",
-                "name": "positive_coords",
-                "shape": 7,
-                "type": "STRING",
-                "link": 267
-              },
-              {
-                "label": "negative_coords",
-                "localized_name": "negative_coords",
-                "name": "negative_coords",
-                "shape": 7,
-                "type": "STRING",
-                "link": 268
-              },
-              {
-                "localized_name": "threshold",
-                "name": "threshold",
-                "type": "FLOAT",
-                "widget": {
-                  "name": "threshold"
-                },
-                "link": 269
-              },
-              {
-                "localized_name": "refine_iterations",
-                "name": "refine_iterations",
-                "type": "INT",
-                "widget": {
-                  "name": "refine_iterations"
-                },
-                "link": 270
-              },
-              {
-                "localized_name": "individual_masks",
-                "name": "individual_masks",
-                "type": "BOOLEAN",
-                "widget": {
-                  "name": "individual_masks"
-                },
-                "link": 271
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "masks",
-                "name": "masks",
-                "type": "MASK",
-                "links": [
-                  231
-                ]
-              },
-              {
-                "localized_name": "bboxes",
-                "name": "bboxes",
-                "type": "BOUNDING_BOX",
-                "links": [
-                  232
-                ]
-              }
-            ],
-            "properties": {
-              "ue_properties": {
-                "widget_ue_connectable": {},
-                "version": "7.7",
-                "input_ue_unconnectable": {}
-              },
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "Node name for S&R": "SAM3_Detect",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65
-            },
-            "widgets_values": [
-              0.5,
-              2,
-              false
-            ]
-          },
-          {
-            "id": 77,
-            "type": "CheckpointLoaderSimple",
-            "pos": [
-              -1970,
-              -3200
-            ],
-            "size": [
-              330,
-              140
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "ckpt_name",
-                "name": "ckpt_name",
-                "type": "COMBO",
-                "widget": {
-                  "name": "ckpt_name"
-                },
-                "link": 272
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "MODEL",
-                "name": "MODEL",
-                "type": "MODEL",
-                "links": [
-                  237
-                ]
-              },
-              {
-                "localized_name": "CLIP",
-                "name": "CLIP",
-                "type": "CLIP",
-                "links": [
-                  240
-                ]
-              },
-              {
-                "localized_name": "VAE",
-                "name": "VAE",
-                "type": "VAE",
-                "links": null
-              }
-            ],
-            "properties": {
-              "ue_properties": {
-                "widget_ue_connectable": {},
-                "version": "7.7",
-                "input_ue_unconnectable": {}
-              },
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "Node name for S&R": "CheckpointLoaderSimple",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "models": [
-                {
-                  "name": "sam3.1_multiplex_fp16.safetensors",
-                  "url": "https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors",
-                  "directory": "checkpoints"
-                }
-              ]
-            },
-            "widgets_values": [
-              "sam3.1_multiplex_fp16.safetensors"
-            ]
-          },
-          {
-            "id": 78,
-            "type": "CLIPTextEncode",
-            "pos": [
-              -2000,
-              -3000
-            ],
-            "size": [
-              400,
-              200
-            ],
-            "flags": {},
-            "order": 2,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "clip",
-                "name": "clip",
-                "type": "CLIP",
-                "link": 240
-              },
-              {
-                "localized_name": "text",
-                "name": "text",
-                "type": "STRING",
-                "widget": {
-                  "name": "text"
-                },
-                "link": 265
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "CONDITIONING",
-                "name": "CONDITIONING",
-                "type": "CONDITIONING",
-                "links": [
-                  200
-                ]
-              }
-            ],
-            "properties": {
-              "ue_properties": {
-                "widget_ue_connectable": {},
-                "version": "7.7",
-                "input_ue_unconnectable": {}
-              },
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "Node name for S&R": "CLIPTextEncode",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65
-            },
-            "widgets_values": [
-              ""
-            ]
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 237,
-            "origin_id": 77,
-            "origin_slot": 0,
-            "target_id": 75,
-            "target_slot": 0,
-            "type": "MODEL"
-          },
-          {
-            "id": 200,
-            "origin_id": 78,
-            "origin_slot": 0,
-            "target_id": 75,
-            "target_slot": 2,
-            "type": "CONDITIONING"
-          },
-          {
-            "id": 240,
-            "origin_id": 77,
-            "origin_slot": 1,
-            "target_id": 78,
-            "target_slot": 0,
-            "type": "CLIP"
-          },
-          {
-            "id": 231,
-            "origin_id": 75,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "MASK"
-          },
-          {
-            "id": 232,
-            "origin_id": 75,
-            "origin_slot": 1,
-            "target_id": -20,
-            "target_slot": 1,
-            "type": "BOUNDING_BOX"
-          },
-          {
-            "id": 264,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 75,
-            "target_slot": 1,
-            "type": "IMAGE"
-          },
-          {
-            "id": 265,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 78,
-            "target_slot": 1,
-            "type": "STRING"
-          },
-          {
-            "id": 266,
-            "origin_id": -10,
-            "origin_slot": 2,
-            "target_id": 75,
-            "target_slot": 3,
-            "type": "BOUNDING_BOX"
-          },
-          {
-            "id": 267,
-            "origin_id": -10,
-            "origin_slot": 3,
-            "target_id": 75,
-            "target_slot": 4,
-            "type": "STRING"
-          },
-          {
-            "id": 268,
-            "origin_id": -10,
-            "origin_slot": 4,
-            "target_id": 75,
-            "target_slot": 5,
-            "type": "STRING"
-          },
-          {
-            "id": 269,
-            "origin_id": -10,
-            "origin_slot": 5,
-            "target_id": 75,
-            "target_slot": 6,
-            "type": "FLOAT"
-          },
-          {
-            "id": 270,
-            "origin_id": -10,
-            "origin_slot": 6,
-            "target_id": 75,
-            "target_slot": 7,
-            "type": "INT"
-          },
-          {
-            "id": 271,
-            "origin_id": -10,
-            "origin_slot": 7,
-            "target_id": 75,
-            "target_slot": 8,
-            "type": "BOOLEAN"
-          },
-          {
-            "id": 272,
-            "origin_id": -10,
-            "origin_slot": 8,
-            "target_id": 77,
-            "target_slot": 0,
-            "type": "COMBO"
-          }
-        ],
-        "extra": {},
-        "category": "Image Tools/Image Segmentation",
-        "description": "Segments images into masks using Meta SAM3 from text prompts, points, or boxes."
-      }
-    ]
-  },
-  "extra": {
-    "ue_links": []
-  }
-}

blueprints/Prompt Enhance.json

@@ -1,279 +1 @@
-{
-  "revision": 0,
-  "last_node_id": 15,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 15,
-      "type": "24d8bbfd-39d4-4774-bff0-3de40cc7a471",
-      "pos": [
-        -1490,
-        2040
-      ],
-      "size": [
-        400,
-        260
-      ],
-      "flags": {},
-      "order": 0,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "prompt",
-          "type": "STRING",
-          "widget": {
-            "name": "prompt"
-          },
-          "link": null
-        },
-        {
-          "label": "reference images",
-          "name": "images",
-          "type": "IMAGE",
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "name": "STRING",
-          "type": "STRING",
-          "links": null
-        }
-      ],
-      "title": "Prompt Enhance",
-      "properties": {
-        "proxyWidgets": [
-          [
-            "-1",
-            "prompt"
-          ]
-        ],
-        "cnr_id": "comfy-core",
-        "ver": "0.14.1"
-      },
-      "widgets_values": [
-        ""
-      ]
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "24d8bbfd-39d4-4774-bff0-3de40cc7a471",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 15,
-          "lastLinkId": 14,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Prompt Enhance",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            -2170,
-            2110,
-            138.876953125,
-            80
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            -640,
-            2110,
-            120,
-            60
-          ]
-        },
-        "inputs": [
-          {
-            "id": "aeab7216-00e0-4528-a09b-bba50845c5a6",
-            "name": "prompt",
-            "type": "STRING",
-            "linkIds": [
-              11
-            ],
-            "pos": [
-              -2051.123046875,
-              2130
-            ]
-          },
-          {
-            "id": "7b73fd36-aa31-4771-9066-f6c83879994b",
-            "name": "images",
-            "type": "IMAGE",
-            "linkIds": [
-              14
-            ],
-            "label": "reference images",
-            "pos": [
-              -2051.123046875,
-              2150
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "c7b0d930-68a1-48d1-b496-0519e5837064",
-            "name": "STRING",
-            "type": "STRING",
-            "linkIds": [
-              13
-            ],
-            "pos": [
-              -620,
-              2130
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 11,
-            "type": "GeminiNode",
-            "pos": [
-              -1560,
-              1990
-            ],
-            "size": [
-              470,
-              470
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "shape": 7,
-                "type": "IMAGE",
-                "link": 14
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "shape": 7,
-                "type": "AUDIO",
-                "link": null
-              },
-              {
-                "localized_name": "video",
-                "name": "video",
-                "shape": 7,
-                "type": "VIDEO",
-                "link": null
-              },
-              {
-                "localized_name": "files",
-                "name": "files",
-                "shape": 7,
-                "type": "GEMINI_INPUT_FILES",
-                "link": null
-              },
-              {
-                "localized_name": "prompt",
-                "name": "prompt",
-                "type": "STRING",
-                "widget": {
-                  "name": "prompt"
-                },
-                "link": 11
-              },
-              {
-                "localized_name": "model",
-                "name": "model",
-                "type": "COMBO",
-                "widget": {
-                  "name": "model"
-                },
-                "link": null
-              },
-              {
-                "localized_name": "seed",
-                "name": "seed",
-                "type": "INT",
-                "widget": {
-                  "name": "seed"
-                },
-                "link": null
-              },
-              {
-                "localized_name": "system_prompt",
-                "name": "system_prompt",
-                "shape": 7,
-                "type": "STRING",
-                "widget": {
-                  "name": "system_prompt"
-                },
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "STRING",
-                "name": "STRING",
-                "type": "STRING",
-                "links": [
-                  13
-                ]
-              }
-            ],
-            "properties": {
-              "cnr_id": "comfy-core",
-              "ver": "0.14.1",
-              "Node name for S&R": "GeminiNode"
-            },
-            "widgets_values": [
-              "",
-              "gemini-3-pro-preview",
-              42,
-              "randomize",
-              "You are an expert in prompt writing.\nBased on the input, rewrite the user's input into a detailed prompt.\nincluding camera settings, lighting, composition, and style.\nReturn the prompt only"
-            ],
-            "color": "#432",
-            "bgcolor": "#653"
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 11,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 11,
-            "target_slot": 4,
-            "type": "STRING"
-          },
-          {
-            "id": 13,
-            "origin_id": 11,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "STRING"
-          },
-          {
-            "id": 14,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 11,
-            "target_slot": 0,
-            "type": "IMAGE"
-          }
-        ],
-        "extra": {
-          "workflowRendererVersion": "LG"
-        },
-        "category": "Text generation/Prompt enhance",
-        "description": "Expands short text prompts into detailed descriptions using a text generation model for better generation quality."
-      }
-    ]
-  },
-  "extra": {}
-}
+{"revision": 0, "last_node_id": 15, "last_link_id": 0, "nodes": [{"id": 15, "type": "24d8bbfd-39d4-4774-bff0-3de40cc7a471", "pos": [-1490, 2040], "size": [400, 260], "flags": {}, "order": 0, "mode": 0, "inputs": [{"name": "prompt", "type": "STRING", "widget": {"name": "prompt"}, "link": null}, {"label": "reference images", "name": "images", "type": "IMAGE", "link": null}], "outputs": [{"name": "STRING", "type": "STRING", "links": null}], "title": "Prompt Enhance", "properties": {"proxyWidgets": [["-1", "prompt"]], "cnr_id": "comfy-core", "ver": "0.14.1"}, "widgets_values": [""]}], "links": [], "version": 0.4, "definitions": {"subgraphs": [{"id": "24d8bbfd-39d4-4774-bff0-3de40cc7a471", "version": 1, "state": {"lastGroupId": 0, "lastNodeId": 15, "lastLinkId": 14, "lastRerouteId": 0}, "revision": 0, "config": {}, "name": "Prompt Enhance", "inputNode": {"id": -10, "bounding": [-2170, 2110, 138.876953125, 80]}, "outputNode": {"id": -20, "bounding": [-640, 2110, 120, 60]}, "inputs": [{"id": "aeab7216-00e0-4528-a09b-bba50845c5a6", "name": "prompt", "type": "STRING", "linkIds": [11], "pos": [-2051.123046875, 2130]}, {"id": "7b73fd36-aa31-4771-9066-f6c83879994b", "name": "images", "type": "IMAGE", "linkIds": [14], "label": "reference images", "pos": [-2051.123046875, 2150]}], "outputs": [{"id": "c7b0d930-68a1-48d1-b496-0519e5837064", "name": "STRING", "type": "STRING", "linkIds": [13], "pos": [-620, 2130]}], "widgets": [], "nodes": [{"id": 11, "type": "GeminiNode", "pos": [-1560, 1990], "size": [470, 470], "flags": {}, "order": 0, "mode": 0, "inputs": [{"localized_name": "images", "name": "images", "shape": 7, "type": "IMAGE", "link": 14}, {"localized_name": "audio", "name": "audio", "shape": 7, "type": "AUDIO", "link": null}, {"localized_name": "video", "name": "video", "shape": 7, "type": "VIDEO", "link": null}, {"localized_name": "files", "name": "files", "shape": 7, "type": "GEMINI_INPUT_FILES", "link": null}, {"localized_name": "prompt", "name": "prompt", "type": "STRING", "widget": {"name": "prompt"}, "link": 11}, {"localized_name": "model", "name": "model", "type": "COMBO", "widget": {"name": "model"}, "link": null}, {"localized_name": "seed", "name": "seed", "type": "INT", "widget": {"name": "seed"}, "link": null}, {"localized_name": "system_prompt", "name": "system_prompt", "shape": 7, "type": "STRING", "widget": {"name": "system_prompt"}, "link": null}], "outputs": [{"localized_name": "STRING", "name": "STRING", "type": "STRING", "links": [13]}], "properties": {"cnr_id": "comfy-core", "ver": "0.14.1", "Node name for S&R": "GeminiNode"}, "widgets_values": ["", "gemini-3-pro-preview", 42, "randomize", "You are an expert in prompt writing.\nBased on the input, rewrite the user's input into a detailed prompt.\nincluding camera settings, lighting, composition, and style.\nReturn the prompt only"], "color": "#432", "bgcolor": "#653"}], "groups": [], "links": [{"id": 11, "origin_id": -10, "origin_slot": 0, "target_id": 11, "target_slot": 4, "type": "STRING"}, {"id": 13, "origin_id": 11, "origin_slot": 0, "target_id": -20, "target_slot": 0, "type": "STRING"}, {"id": 14, "origin_id": -10, "origin_slot": 1, "target_id": 11, "target_slot": 0, "type": "IMAGE"}], "extra": {"workflowRendererVersion": "LG"}, "category": "Text generation/Prompt enhance"}]}, "extra": {}}

blueprints/Remove Background (BiRefNet).json

@@ -1,397 +0,0 @@
-{
-  "revision": 0,
-  "last_node_id": 19,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 19,
-      "type": "5b40ca21-ba1a-41d5-b403-4d2d7acdc195",
-      "pos": [
-        -6411.330578108367,
-        1940.2638932730042
-      ],
-      "size": [
-        349.609375,
-        145.9375
-      ],
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "inputs": [
-        {
-          "localized_name": "image",
-          "name": "image",
-          "type": "IMAGE",
-          "link": null
-        },
-        {
-          "name": "bg_removal_name",
-          "type": "COMBO",
-          "widget": {
-            "name": "bg_removal_name"
-          },
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "localized_name": "IMAGE",
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": []
-        },
-        {
-          "name": "mask",
-          "type": "MASK",
-          "links": []
-        }
-      ],
-      "properties": {
-        "proxyWidgets": [
-          [
-            "14",
-            "bg_removal_name"
-          ]
-        ]
-      },
-      "widgets_values": [],
-      "title": "Remove Background (BiRefNet)"
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "5b40ca21-ba1a-41d5-b403-4d2d7acdc195",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 21,
-          "lastLinkId": 16,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Remove Background (BiRefNet)",
-        "description": "Removes or replaces image backgrounds using BiRefNet segmentation and alpha compositing.",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            -6728.534070722246,
-            1475.2619799128663,
-            150.9140625,
-            88
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            -6169.049695722246,
-            1475.2619799128663,
-            128,
-            88
-          ]
-        },
-        "inputs": [
-          {
-            "id": "7bc321cd-df31-4c39-aaf7-7f0d01326189",
-            "name": "image",
-            "type": "IMAGE",
-            "linkIds": [
-              5,
-              7
-            ],
-            "localized_name": "image",
-            "pos": [
-              -6601.620008222246,
-              1499.2619799128663
-            ]
-          },
-          {
-            "id": "e89d2cd8-daa3-4e29-8a69-851db85072cb",
-            "name": "bg_removal_name",
-            "type": "COMBO",
-            "linkIds": [
-              12
-            ],
-            "pos": [
-              -6601.620008222246,
-              1519.2619799128663
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "16e7863c-4c38-46c2-aa74-e82991fbfe8d",
-            "name": "IMAGE",
-            "type": "IMAGE",
-            "linkIds": [
-              8
-            ],
-            "localized_name": "IMAGE",
-            "pos": [
-              -6145.049695722246,
-              1499.2619799128663
-            ]
-          },
-          {
-            "id": "f7240c19-5b80-406e-a8e2-9b12440ee2d6",
-            "name": "mask",
-            "type": "MASK",
-            "linkIds": [
-              11
-            ],
-            "pos": [
-              -6145.049695722246,
-              1519.2619799128663
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 13,
-            "type": "RemoveBackground",
-            "pos": [
-              -6536.764823982709,
-              1444.9963409012412
-            ],
-            "size": [
-              302.25,
-              72
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 5
-              },
-              {
-                "localized_name": "bg_removal_model",
-                "name": "bg_removal_model",
-                "type": "BACKGROUND_REMOVAL",
-                "link": 3
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "mask",
-                "name": "mask",
-                "type": "MASK",
-                "links": [
-                  4,
-                  11
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "RemoveBackground"
-            }
-          },
-          {
-            "id": 14,
-            "type": "LoadBackgroundRemovalModel",
-            "pos": [
-              -6540.534070722246,
-              1302.223464635445
-            ],
-            "size": [
-              311.484375,
-              85.515625
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "bg_removal_name",
-                "name": "bg_removal_name",
-                "type": "COMBO",
-                "widget": {
-                  "name": "bg_removal_name"
-                },
-                "link": 12
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "bg_model",
-                "name": "bg_model",
-                "type": "BACKGROUND_REMOVAL",
-                "links": [
-                  3
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "LoadBackgroundRemovalModel",
-              "models": [
-                {
-                  "name": "birefnet.safetensors",
-                  "url": "https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors",
-                  "directory": "background_removal"
-                }
-              ]
-            },
-            "widgets_values": [
-              "birefnet.safetensors"
-            ]
-          },
-          {
-            "id": 15,
-            "type": "InvertMask",
-            "pos": [
-              -6532.446160529669,
-              1571.1111286839914
-            ],
-            "size": [
-              285.984375,
-              48
-            ],
-            "flags": {},
-            "order": 2,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "mask",
-                "name": "mask",
-                "type": "MASK",
-                "link": 4
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "MASK",
-                "name": "MASK",
-                "type": "MASK",
-                "links": [
-                  6
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "InvertMask"
-            }
-          },
-          {
-            "id": 16,
-            "type": "JoinImageWithAlpha",
-            "pos": [
-              -6527.4370171636665,
-              1674.3004951902876
-            ],
-            "size": [
-              284.96875,
-              72
-            ],
-            "flags": {},
-            "order": 3,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 7
-              },
-              {
-                "localized_name": "alpha",
-                "name": "alpha",
-                "type": "MASK",
-                "link": 6
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "IMAGE",
-                "name": "IMAGE",
-                "type": "IMAGE",
-                "links": [
-                  8
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "JoinImageWithAlpha"
-            }
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 3,
-            "origin_id": 14,
-            "origin_slot": 0,
-            "target_id": 13,
-            "target_slot": 1,
-            "type": "BACKGROUND_REMOVAL"
-          },
-          {
-            "id": 4,
-            "origin_id": 13,
-            "origin_slot": 0,
-            "target_id": 15,
-            "target_slot": 0,
-            "type": "MASK"
-          },
-          {
-            "id": 6,
-            "origin_id": 15,
-            "origin_slot": 0,
-            "target_id": 16,
-            "target_slot": 1,
-            "type": "MASK"
-          },
-          {
-            "id": 5,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 13,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 7,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 16,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 8,
-            "origin_id": 16,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 11,
-            "origin_id": 13,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 1,
-            "type": "MASK"
-          },
-          {
-            "id": 12,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 14,
-            "target_slot": 0,
-            "type": "COMBO"
-          }
-        ],
-        "extra": {},
-        "category": "Image generation and editing/Background Removal"
-      }
-    ]
-  },
-  "extra": {}
-}

blueprints/Sharpen.json

@@ -1,310 +1 @@
-{
-  "revision": 0,
-  "last_node_id": 25,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 25,
-      "type": "621ba4e2-22a8-482d-a369-023753198b7b",
-      "pos": [
-        4610,
-        -790
-      ],
-      "size": [
-        230,
-        58
-      ],
-      "flags": {},
-      "order": 4,
-      "mode": 0,
-      "inputs": [
-        {
-          "label": "image",
-          "localized_name": "images.image0",
-          "name": "images.image0",
-          "type": "IMAGE",
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "label": "IMAGE",
-          "localized_name": "IMAGE0",
-          "name": "IMAGE0",
-          "type": "IMAGE",
-          "links": []
-        }
-      ],
-      "title": "Sharpen",
-      "properties": {
-        "proxyWidgets": [
-          [
-            "24",
-            "value"
-          ]
-        ]
-      },
-      "widgets_values": []
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "621ba4e2-22a8-482d-a369-023753198b7b",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 24,
-          "lastLinkId": 36,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Sharpen",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            4090,
-            -825,
-            120,
-            60
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            5150,
-            -825,
-            120,
-            60
-          ]
-        },
-        "inputs": [
-          {
-            "id": "37011fb7-14b7-4e0e-b1a0-6a02e8da1fd7",
-            "name": "images.image0",
-            "type": "IMAGE",
-            "linkIds": [
-              34
-            ],
-            "localized_name": "images.image0",
-            "label": "image",
-            "pos": [
-              4190,
-              -805
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "e9182b3f-635c-4cd4-a152-4b4be17ae4b9",
-            "name": "IMAGE0",
-            "type": "IMAGE",
-            "linkIds": [
-              35
-            ],
-            "localized_name": "IMAGE0",
-            "label": "IMAGE",
-            "pos": [
-              5170,
-              -805
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 24,
-            "type": "PrimitiveFloat",
-            "pos": [
-              4280,
-              -1240
-            ],
-            "size": [
-              270,
-              58
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "strength",
-                "localized_name": "value",
-                "name": "value",
-                "type": "FLOAT",
-                "widget": {
-                  "name": "value"
-                },
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "FLOAT",
-                "name": "FLOAT",
-                "type": "FLOAT",
-                "links": [
-                  36
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "PrimitiveFloat",
-              "min": 0,
-              "max": 3,
-              "precision": 2,
-              "step": 0.05
-            },
-            "widgets_values": [
-              0.5
-            ]
-          },
-          {
-            "id": 23,
-            "type": "GLSLShader",
-            "pos": [
-              4570,
-              -1240
-            ],
-            "size": [
-              370,
-              192
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "image0",
-                "localized_name": "images.image0",
-                "name": "images.image0",
-                "type": "IMAGE",
-                "link": 34
-              },
-              {
-                "label": "image1",
-                "localized_name": "images.image1",
-                "name": "images.image1",
-                "shape": 7,
-                "type": "IMAGE",
-                "link": null
-              },
-              {
-                "label": "u_float0",
-                "localized_name": "floats.u_float0",
-                "name": "floats.u_float0",
-                "shape": 7,
-                "type": "FLOAT",
-                "link": 36
-              },
-              {
-                "label": "u_float1",
-                "localized_name": "floats.u_float1",
-                "name": "floats.u_float1",
-                "shape": 7,
-                "type": "FLOAT",
-                "link": null
-              },
-              {
-                "label": "u_int0",
-                "localized_name": "ints.u_int0",
-                "name": "ints.u_int0",
-                "shape": 7,
-                "type": "INT",
-                "link": null
-              },
-              {
-                "localized_name": "fragment_shader",
-                "name": "fragment_shader",
-                "type": "STRING",
-                "widget": {
-                  "name": "fragment_shader"
-                },
-                "link": null
-              },
-              {
-                "localized_name": "size_mode",
-                "name": "size_mode",
-                "type": "COMFY_DYNAMICCOMBO_V3",
-                "widget": {
-                  "name": "size_mode"
-                },
-                "link": null
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "IMAGE0",
-                "name": "IMAGE0",
-                "type": "IMAGE",
-                "links": [
-                  35
-                ]
-              },
-              {
-                "localized_name": "IMAGE1",
-                "name": "IMAGE1",
-                "type": "IMAGE",
-                "links": null
-              },
-              {
-                "localized_name": "IMAGE2",
-                "name": "IMAGE2",
-                "type": "IMAGE",
-                "links": null
-              },
-              {
-                "localized_name": "IMAGE3",
-                "name": "IMAGE3",
-                "type": "IMAGE",
-                "links": null
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "GLSLShader"
-            },
-            "widgets_values": [
-              "#version 300 es\nprecision highp float;\n\nuniform sampler2D u_image0;\nuniform float u_float0;  // strength [0.0 – 2.0] typical: 0.3–1.0\n\nin vec2 v_texCoord;\nlayout(location = 0) out vec4 fragColor0;\n\nvoid main() {\n    vec2 texel = 1.0 / vec2(textureSize(u_image0, 0));\n    \n    // Sample center and neighbors\n    vec4 center = texture(u_image0, v_texCoord);\n    vec4 top    = texture(u_image0, v_texCoord + vec2( 0.0, -texel.y));\n    vec4 bottom = texture(u_image0, v_texCoord + vec2( 0.0,  texel.y));\n    vec4 left   = texture(u_image0, v_texCoord + vec2(-texel.x,  0.0));\n    vec4 right  = texture(u_image0, v_texCoord + vec2( texel.x,  0.0));\n    \n    // Edge enhancement (Laplacian)\n    vec4 edges = center * 4.0 - top - bottom - left - right;\n    \n    // Add edges back scaled by strength\n    vec4 sharpened = center + edges * u_float0;\n    \n    fragColor0 = vec4(clamp(sharpened.rgb, 0.0, 1.0), center.a);\n}",
-              "from_input"
-            ]
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 36,
-            "origin_id": 24,
-            "origin_slot": 0,
-            "target_id": 23,
-            "target_slot": 2,
-            "type": "FLOAT"
-          },
-          {
-            "id": 34,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 23,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 35,
-            "origin_id": 23,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "IMAGE"
-          }
-        ],
-        "extra": {
-          "workflowRendererVersion": "LG"
-        },
-        "category": "Image Tools/Sharpen",
-        "description": "Sharpens image details using a GPU fragment shader for enhanced clarity."
-      }
-    ]
-  }
-}
+{"revision": 0, "last_node_id": 25, "last_link_id": 0, "nodes": [{"id": 25, "type": "621ba4e2-22a8-482d-a369-023753198b7b", "pos": [4610, -790], "size": [230, 58], "flags": {}, "order": 4, "mode": 0, "inputs": [{"label": "image", "localized_name": "images.image0", "name": "images.image0", "type": "IMAGE", "link": null}], "outputs": [{"label": "IMAGE", "localized_name": "IMAGE0", "name": "IMAGE0", "type": "IMAGE", "links": []}], "title": "Sharpen", "properties": {"proxyWidgets": [["24", "value"]]}, "widgets_values": []}], "links": [], "version": 0.4, "definitions": {"subgraphs": [{"id": "621ba4e2-22a8-482d-a369-023753198b7b", "version": 1, "state": {"lastGroupId": 0, "lastNodeId": 24, "lastLinkId": 36, "lastRerouteId": 0}, "revision": 0, "config": {}, "name": "Sharpen", "inputNode": {"id": -10, "bounding": [4090, -825, 120, 60]}, "outputNode": {"id": -20, "bounding": [5150, -825, 120, 60]}, "inputs": [{"id": "37011fb7-14b7-4e0e-b1a0-6a02e8da1fd7", "name": "images.image0", "type": "IMAGE", "linkIds": [34], "localized_name": "images.image0", "label": "image", "pos": [4190, -805]}], "outputs": [{"id": "e9182b3f-635c-4cd4-a152-4b4be17ae4b9", "name": "IMAGE0", "type": "IMAGE", "linkIds": [35], "localized_name": "IMAGE0", "label": "IMAGE", "pos": [5170, -805]}], "widgets": [], "nodes": [{"id": 24, "type": "PrimitiveFloat", "pos": [4280, -1240], "size": [270, 58], "flags": {}, "order": 0, "mode": 0, "inputs": [{"label": "strength", "localized_name": "value", "name": "value", "type": "FLOAT", "widget": {"name": "value"}, "link": null}], "outputs": [{"localized_name": "FLOAT", "name": "FLOAT", "type": "FLOAT", "links": [36]}], "properties": {"Node name for S&R": "PrimitiveFloat", "min": 0, "max": 3, "precision": 2, "step": 0.05}, "widgets_values": [0.5]}, {"id": 23, "type": "GLSLShader", "pos": [4570, -1240], "size": [370, 192], "flags": {}, "order": 1, "mode": 0, "inputs": [{"label": "image0", "localized_name": "images.image0", "name": "images.image0", "type": "IMAGE", "link": 34}, {"label": "image1", "localized_name": "images.image1", "name": "images.image1", "shape": 7, "type": "IMAGE", "link": null}, {"label": "u_float0", "localized_name": "floats.u_float0", "name": "floats.u_float0", "shape": 7, "type": "FLOAT", "link": 36}, {"label": "u_float1", "localized_name": "floats.u_float1", "name": "floats.u_float1", "shape": 7, "type": "FLOAT", "link": null}, {"label": "u_int0", "localized_name": "ints.u_int0", "name": "ints.u_int0", "shape": 7, "type": "INT", "link": null}, {"localized_name": "fragment_shader", "name": "fragment_shader", "type": "STRING", "widget": {"name": "fragment_shader"}, "link": null}, {"localized_name": "size_mode", "name": "size_mode", "type": "COMFY_DYNAMICCOMBO_V3", "widget": {"name": "size_mode"}, "link": null}], "outputs": [{"localized_name": "IMAGE0", "name": "IMAGE0", "type": "IMAGE", "links": [35]}, {"localized_name": "IMAGE1", "name": "IMAGE1", "type": "IMAGE", "links": null}, {"localized_name": "IMAGE2", "name": "IMAGE2", "type": "IMAGE", "links": null}, {"localized_name": "IMAGE3", "name": "IMAGE3", "type": "IMAGE", "links": null}], "properties": {"Node name for S&R": "GLSLShader"}, "widgets_values": ["#version 300 es\nprecision highp float;\n\nuniform sampler2D u_image0;\nuniform vec2 u_resolution;\nuniform float u_float0;  // strength [0.0 – 2.0] typical: 0.3–1.0\n\nin vec2 v_texCoord;\nlayout(location = 0) out vec4 fragColor0;\n\nvoid main() {\n    vec2 texel = 1.0 / u_resolution;\n    \n    // Sample center and neighbors\n    vec4 center = texture(u_image0, v_texCoord);\n    vec4 top    = texture(u_image0, v_texCoord + vec2( 0.0, -texel.y));\n    vec4 bottom = texture(u_image0, v_texCoord + vec2( 0.0,  texel.y));\n    vec4 left   = texture(u_image0, v_texCoord + vec2(-texel.x,  0.0));\n    vec4 right  = texture(u_image0, v_texCoord + vec2( texel.x,  0.0));\n    \n    // Edge enhancement (Laplacian)\n    vec4 edges = center * 4.0 - top - bottom - left - right;\n    \n    // Add edges back scaled by strength\n    vec4 sharpened = center + edges * u_float0;\n    \n    fragColor0 = vec4(clamp(sharpened.rgb, 0.0, 1.0), center.a);\n}", "from_input"]}], "groups": [], "links": [{"id": 36, "origin_id": 24, "origin_slot": 0, "target_id": 23, "target_slot": 2, "type": "FLOAT"}, {"id": 34, "origin_id": -10, "origin_slot": 0, "target_id": 23, "target_slot": 0, "type": "IMAGE"}, {"id": 35, "origin_id": 23, "origin_slot": 0, "target_id": -20, "target_slot": 0, "type": "IMAGE"}], "extra": {"workflowRendererVersion": "LG"}, "category": "Image Tools/Sharpen"}]}}

blueprints/Video Segmentation (SAM3).json

@@ -1,827 +0,0 @@
-{
-  "revision": 0,
-  "last_node_id": 130,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 130,
-      "type": "7937cf78-b52b-40a3-93b2-b4e2e5f98df1",
-      "pos": [
-        -1210,
-        -2780
-      ],
-      "size": [
-        300,
-        370
-      ],
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "video",
-          "type": "VIDEO",
-          "link": null
-        },
-        {
-          "name": "text",
-          "type": "STRING",
-          "widget": {
-            "name": "text"
-          },
-          "link": null
-        },
-        {
-          "name": "bboxes",
-          "type": "BOUNDING_BOX",
-          "link": null
-        },
-        {
-          "name": "positive_coords",
-          "type": "STRING",
-          "link": null
-        },
-        {
-          "name": "negative_coords",
-          "type": "STRING",
-          "link": null
-        },
-        {
-          "name": "threshold",
-          "type": "FLOAT",
-          "widget": {
-            "name": "threshold"
-          },
-          "link": null
-        },
-        {
-          "name": "refine_iterations",
-          "type": "INT",
-          "widget": {
-            "name": "refine_iterations"
-          },
-          "link": null
-        },
-        {
-          "name": "individual_masks",
-          "type": "BOOLEAN",
-          "widget": {
-            "name": "individual_masks"
-          },
-          "link": null
-        },
-        {
-          "name": "ckpt_name",
-          "type": "COMBO",
-          "widget": {
-            "name": "ckpt_name"
-          },
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "localized_name": "masks",
-          "name": "masks",
-          "type": "MASK",
-          "links": []
-        },
-        {
-          "localized_name": "bboxes",
-          "name": "bboxes",
-          "type": "BOUNDING_BOX",
-          "links": []
-        },
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "links": null
-        },
-        {
-          "name": "fps",
-          "type": "FLOAT",
-          "links": null
-        }
-      ],
-      "properties": {
-        "proxyWidgets": [
-          [
-            "125",
-            "text"
-          ],
-          [
-            "126",
-            "threshold"
-          ],
-          [
-            "126",
-            "refine_iterations"
-          ],
-          [
-            "126",
-            "individual_masks"
-          ],
-          [
-            "127",
-            "ckpt_name"
-          ]
-        ],
-        "cnr_id": "comfy-core",
-        "ver": "0.19.3",
-        "enableTabs": false,
-        "tabWidth": 65,
-        "tabXOffset": 10,
-        "hasSecondTab": false,
-        "secondTabText": "Send Back",
-        "secondTabOffset": 80,
-        "secondTabWidth": 65
-      },
-      "widgets_values": [],
-      "title": "Video Segmentation (SAM3)"
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "7937cf78-b52b-40a3-93b2-b4e2e5f98df1",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 130,
-          "lastLinkId": 299,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Video Segmentation (SAM3)",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            -2260,
-            -3450,
-            136.369140625,
-            220
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            -1050,
-            -3510,
-            120,
-            120
-          ]
-        },
-        "inputs": [
-          {
-            "id": "680ffd88-32fe-48be-88d6-91ea44d5eaee",
-            "name": "video",
-            "type": "VIDEO",
-            "linkIds": [
-              252
-            ],
-            "pos": [
-              -2143.630859375,
-              -3430
-            ]
-          },
-          {
-            "id": "ceaf249c-32d7-4624-8bf6-e590e347ed90",
-            "name": "text",
-            "type": "STRING",
-            "linkIds": [
-              254
-            ],
-            "pos": [
-              -2143.630859375,
-              -3410
-            ]
-          },
-          {
-            "id": "1ffbff36-da0c-4854-8cb4-88ad31e64f99",
-            "name": "bboxes",
-            "type": "BOUNDING_BOX",
-            "linkIds": [
-              255
-            ],
-            "pos": [
-              -2143.630859375,
-              -3390
-            ]
-          },
-          {
-            "id": "67b7f4c7-cec0-4e00-b154-23cc1abf880e",
-            "name": "positive_coords",
-            "type": "STRING",
-            "linkIds": [
-              256
-            ],
-            "pos": [
-              -2143.630859375,
-              -3370
-            ]
-          },
-          {
-            "id": "b090a498-2bde-46b9-9554-18501401d687",
-            "name": "negative_coords",
-            "type": "STRING",
-            "linkIds": [
-              257
-            ],
-            "pos": [
-              -2143.630859375,
-              -3350
-            ]
-          },
-          {
-            "id": "1a76dfcf-ce95-46af-bba5-c42160c683dd",
-            "name": "threshold",
-            "type": "FLOAT",
-            "linkIds": [
-              261
-            ],
-            "pos": [
-              -2143.630859375,
-              -3330
-            ]
-          },
-          {
-            "id": "999523fa-c476-4c53-80c3-0a2f554d18ab",
-            "name": "refine_iterations",
-            "type": "INT",
-            "linkIds": [
-              262
-            ],
-            "pos": [
-              -2143.630859375,
-              -3310
-            ]
-          },
-          {
-            "id": "d2371011-7fe5-4a39-b0c1-df2e0bbd6ece",
-            "name": "individual_masks",
-            "type": "BOOLEAN",
-            "linkIds": [
-              263
-            ],
-            "pos": [
-              -2143.630859375,
-              -3290
-            ]
-          },
-          {
-            "id": "675a8b37-17db-48d1-853c-2fe5d6a74582",
-            "name": "ckpt_name",
-            "type": "COMBO",
-            "linkIds": [
-              273
-            ],
-            "pos": [
-              -2143.630859375,
-              -3270
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "ff50da09-1e59-4a58-9b7f-be1a00aa5913",
-            "name": "masks",
-            "type": "MASK",
-            "linkIds": [
-              231
-            ],
-            "localized_name": "masks",
-            "pos": [
-              -1030,
-              -3490
-            ]
-          },
-          {
-            "id": "8f622e40-8528-4078-b7d3-147e9f872194",
-            "name": "bboxes",
-            "type": "BOUNDING_BOX",
-            "linkIds": [
-              232
-            ],
-            "localized_name": "bboxes",
-            "pos": [
-              -1030,
-              -3470
-            ]
-          },
-          {
-            "id": "6c9924ec-f0fa-4509-83ea-8f97f5889bcc",
-            "name": "audio",
-            "type": "AUDIO",
-            "linkIds": [
-              259
-            ],
-            "pos": [
-              -1030,
-              -3450
-            ]
-          },
-          {
-            "id": "82c1cddc-ab11-44eb-9e2f-1a5c7ea5645b",
-            "name": "fps",
-            "type": "FLOAT",
-            "linkIds": [
-              260
-            ],
-            "pos": [
-              -1030,
-              -3430
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 125,
-            "type": "CLIPTextEncode",
-            "pos": [
-              -2010,
-              -3040
-            ],
-            "size": [
-              400,
-              200
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "clip",
-                "name": "clip",
-                "type": "CLIP",
-                "link": 240
-              },
-              {
-                "localized_name": "text",
-                "name": "text",
-                "type": "STRING",
-                "widget": {
-                  "name": "text"
-                },
-                "link": 254
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "CONDITIONING",
-                "name": "CONDITIONING",
-                "type": "CONDITIONING",
-                "links": [
-                  200
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "CLIPTextEncode",
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65
-            },
-            "widgets_values": [
-              ""
-            ]
-          },
-          {
-            "id": 126,
-            "type": "SAM3_Detect",
-            "pos": [
-              -1520,
-              -3520
-            ],
-            "size": [
-              270,
-              290
-            ],
-            "flags": {},
-            "order": 2,
-            "mode": 0,
-            "inputs": [
-              {
-                "label": "model",
-                "localized_name": "model",
-                "name": "model",
-                "type": "MODEL",
-                "link": 237
-              },
-              {
-                "label": "image",
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 253
-              },
-              {
-                "label": "conditioning",
-                "localized_name": "conditioning",
-                "name": "conditioning",
-                "shape": 7,
-                "type": "CONDITIONING",
-                "link": 200
-              },
-              {
-                "label": "bboxes",
-                "localized_name": "bboxes",
-                "name": "bboxes",
-                "shape": 7,
-                "type": "BOUNDING_BOX",
-                "link": 255
-              },
-              {
-                "label": "positive_coords",
-                "localized_name": "positive_coords",
-                "name": "positive_coords",
-                "shape": 7,
-                "type": "STRING",
-                "link": 256
-              },
-              {
-                "label": "negative_coords",
-                "localized_name": "negative_coords",
-                "name": "negative_coords",
-                "shape": 7,
-                "type": "STRING",
-                "link": 257
-              },
-              {
-                "localized_name": "threshold",
-                "name": "threshold",
-                "type": "FLOAT",
-                "widget": {
-                  "name": "threshold"
-                },
-                "link": 261
-              },
-              {
-                "localized_name": "refine_iterations",
-                "name": "refine_iterations",
-                "type": "INT",
-                "widget": {
-                  "name": "refine_iterations"
-                },
-                "link": 262
-              },
-              {
-                "localized_name": "individual_masks",
-                "name": "individual_masks",
-                "type": "BOOLEAN",
-                "widget": {
-                  "name": "individual_masks"
-                },
-                "link": 263
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "masks",
-                "name": "masks",
-                "type": "MASK",
-                "links": [
-                  231
-                ]
-              },
-              {
-                "localized_name": "bboxes",
-                "name": "bboxes",
-                "type": "BOUNDING_BOX",
-                "links": [
-                  232
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "SAM3_Detect",
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65
-            },
-            "widgets_values": [
-              0.5,
-              2,
-              false
-            ]
-          },
-          {
-            "id": 127,
-            "type": "CheckpointLoaderSimple",
-            "pos": [
-              -1970,
-              -3310
-            ],
-            "size": [
-              330,
-              160
-            ],
-            "flags": {},
-            "order": 3,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "ckpt_name",
-                "name": "ckpt_name",
-                "type": "COMBO",
-                "widget": {
-                  "name": "ckpt_name"
-                },
-                "link": 273
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "MODEL",
-                "name": "MODEL",
-                "type": "MODEL",
-                "links": [
-                  237
-                ]
-              },
-              {
-                "localized_name": "CLIP",
-                "name": "CLIP",
-                "type": "CLIP",
-                "links": [
-                  240
-                ]
-              },
-              {
-                "localized_name": "VAE",
-                "name": "VAE",
-                "type": "VAE",
-                "links": null
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "CheckpointLoaderSimple",
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65,
-              "models": [
-                {
-                  "name": "sam3.1_multiplex_fp16.safetensors",
-                  "url": "https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors",
-                  "directory": "checkpoints"
-                }
-              ]
-            },
-            "widgets_values": [
-              "sam3.1_multiplex_fp16.safetensors"
-            ]
-          },
-          {
-            "id": 128,
-            "type": "GetVideoComponents",
-            "pos": [
-              -1910,
-              -3540
-            ],
-            "size": [
-              230,
-              120
-            ],
-            "flags": {},
-            "order": 4,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "video",
-                "name": "video",
-                "type": "VIDEO",
-                "link": 252
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "links": [
-                  253
-                ]
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "type": "AUDIO",
-                "links": [
-                  259
-                ]
-              },
-              {
-                "localized_name": "fps",
-                "name": "fps",
-                "type": "FLOAT",
-                "links": [
-                  260
-                ]
-              }
-            ],
-            "properties": {
-              "Node name for S&R": "GetVideoComponents",
-              "cnr_id": "comfy-core",
-              "ver": "0.19.3",
-              "enableTabs": false,
-              "tabWidth": 65,
-              "tabXOffset": 10,
-              "hasSecondTab": false,
-              "secondTabText": "Send Back",
-              "secondTabOffset": 80,
-              "secondTabWidth": 65
-            }
-          },
-          {
-            "id": 129,
-            "type": "Note",
-            "pos": [
-              -1980,
-              -2790
-            ],
-            "size": [
-              370,
-              250
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [],
-            "outputs": [],
-            "title": "Note: Prompt format",
-            "properties": {},
-            "widgets_values": [
-              "Max tokens for this model is only 32, to separately prompt multiple subjects you can separate prompts with comma, and set the max amount of objects detected for each prompt with :N\n\nFor example above test prompt finds 2 cakes, one apron, 4 window panels"
-            ],
-            "color": "#432",
-            "bgcolor": "#653"
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 237,
-            "origin_id": 127,
-            "origin_slot": 0,
-            "target_id": 126,
-            "target_slot": 0,
-            "type": "MODEL"
-          },
-          {
-            "id": 200,
-            "origin_id": 125,
-            "origin_slot": 0,
-            "target_id": 126,
-            "target_slot": 2,
-            "type": "CONDITIONING"
-          },
-          {
-            "id": 240,
-            "origin_id": 127,
-            "origin_slot": 1,
-            "target_id": 125,
-            "target_slot": 0,
-            "type": "CLIP"
-          },
-          {
-            "id": 231,
-            "origin_id": 126,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "MASK"
-          },
-          {
-            "id": 232,
-            "origin_id": 126,
-            "origin_slot": 1,
-            "target_id": -20,
-            "target_slot": 1,
-            "type": "BOUNDING_BOX"
-          },
-          {
-            "id": 252,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 128,
-            "target_slot": 0,
-            "type": "VIDEO"
-          },
-          {
-            "id": 253,
-            "origin_id": 128,
-            "origin_slot": 0,
-            "target_id": 126,
-            "target_slot": 1,
-            "type": "IMAGE"
-          },
-          {
-            "id": 254,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 125,
-            "target_slot": 1,
-            "type": "STRING"
-          },
-          {
-            "id": 255,
-            "origin_id": -10,
-            "origin_slot": 2,
-            "target_id": 126,
-            "target_slot": 3,
-            "type": "BOUNDING_BOX"
-          },
-          {
-            "id": 256,
-            "origin_id": -10,
-            "origin_slot": 3,
-            "target_id": 126,
-            "target_slot": 4,
-            "type": "STRING"
-          },
-          {
-            "id": 257,
-            "origin_id": -10,
-            "origin_slot": 4,
-            "target_id": 126,
-            "target_slot": 5,
-            "type": "STRING"
-          },
-          {
-            "id": 259,
-            "origin_id": 128,
-            "origin_slot": 1,
-            "target_id": -20,
-            "target_slot": 2,
-            "type": "AUDIO"
-          },
-          {
-            "id": 260,
-            "origin_id": 128,
-            "origin_slot": 2,
-            "target_id": -20,
-            "target_slot": 3,
-            "type": "FLOAT"
-          },
-          {
-            "id": 261,
-            "origin_id": -10,
-            "origin_slot": 5,
-            "target_id": 126,
-            "target_slot": 6,
-            "type": "FLOAT"
-          },
-          {
-            "id": 262,
-            "origin_id": -10,
-            "origin_slot": 6,
-            "target_id": 126,
-            "target_slot": 7,
-            "type": "INT"
-          },
-          {
-            "id": 263,
-            "origin_id": -10,
-            "origin_slot": 7,
-            "target_id": 126,
-            "target_slot": 8,
-            "type": "BOOLEAN"
-          },
-          {
-            "id": 273,
-            "origin_id": -10,
-            "origin_slot": 8,
-            "target_id": 127,
-            "target_slot": 0,
-            "type": "COMBO"
-          }
-        ],
-        "extra": {},
-        "category": "Video Tools",
-        "description": "Segments video into temporally consistent masks using Meta SAM3 from text or interactive prompts."
-      }
-    ]
-  },
-  "extra": {}
-}

blueprints/Video Upscale(GAN x4).json

@@ -1,421 +1 @@
-{
-  "revision": 0,
-  "last_node_id": 13,
-  "last_link_id": 0,
-  "nodes": [
-    {
-      "id": 13,
-      "type": "cf95b747-3e17-46cb-8097-cac60ff9b2e1",
-      "pos": [
-        1120,
-        330
-      ],
-      "size": [
-        240,
-        58
-      ],
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [
-        {
-          "localized_name": "video",
-          "name": "video",
-          "type": "VIDEO",
-          "link": null
-        },
-        {
-          "name": "model_name",
-          "type": "COMBO",
-          "widget": {
-            "name": "model_name"
-          },
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "localized_name": "VIDEO",
-          "name": "VIDEO",
-          "type": "VIDEO",
-          "links": []
-        }
-      ],
-      "title": "Video Upscale(GAN x4)",
-      "properties": {
-        "proxyWidgets": [
-          [
-            "-1",
-            "model_name"
-          ]
-        ],
-        "cnr_id": "comfy-core",
-        "ver": "0.14.1"
-      },
-      "widgets_values": [
-        "RealESRGAN_x4plus.safetensors"
-      ]
-    }
-  ],
-  "links": [],
-  "version": 0.4,
-  "definitions": {
-    "subgraphs": [
-      {
-        "id": "cf95b747-3e17-46cb-8097-cac60ff9b2e1",
-        "version": 1,
-        "state": {
-          "lastGroupId": 0,
-          "lastNodeId": 13,
-          "lastLinkId": 19,
-          "lastRerouteId": 0
-        },
-        "revision": 0,
-        "config": {},
-        "name": "Video Upscale(GAN x4)",
-        "inputNode": {
-          "id": -10,
-          "bounding": [
-            550,
-            460,
-            120,
-            80
-          ]
-        },
-        "outputNode": {
-          "id": -20,
-          "bounding": [
-            1490,
-            460,
-            120,
-            60
-          ]
-        },
-        "inputs": [
-          {
-            "id": "666d633e-93e7-42dc-8d11-2b7b99b0f2a6",
-            "name": "video",
-            "type": "VIDEO",
-            "linkIds": [
-              10
-            ],
-            "localized_name": "video",
-            "pos": [
-              650,
-              480
-            ]
-          },
-          {
-            "id": "2e23a087-caa8-4d65-99e6-662761aa905a",
-            "name": "model_name",
-            "type": "COMBO",
-            "linkIds": [
-              19
-            ],
-            "pos": [
-              650,
-              500
-            ]
-          }
-        ],
-        "outputs": [
-          {
-            "id": "0c1768ea-3ec2-412f-9af6-8e0fa36dae70",
-            "name": "VIDEO",
-            "type": "VIDEO",
-            "linkIds": [
-              15
-            ],
-            "localized_name": "VIDEO",
-            "pos": [
-              1510,
-              480
-            ]
-          }
-        ],
-        "widgets": [],
-        "nodes": [
-          {
-            "id": 2,
-            "type": "ImageUpscaleWithModel",
-            "pos": [
-              1110,
-              450
-            ],
-            "size": [
-              320,
-              46
-            ],
-            "flags": {},
-            "order": 1,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "upscale_model",
-                "name": "upscale_model",
-                "type": "UPSCALE_MODEL",
-                "link": 1
-              },
-              {
-                "localized_name": "image",
-                "name": "image",
-                "type": "IMAGE",
-                "link": 14
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "IMAGE",
-                "name": "IMAGE",
-                "type": "IMAGE",
-                "links": [
-                  13
-                ]
-              }
-            ],
-            "properties": {
-              "cnr_id": "comfy-core",
-              "ver": "0.10.0",
-              "Node name for S&R": "ImageUpscaleWithModel"
-            }
-          },
-          {
-            "id": 11,
-            "type": "CreateVideo",
-            "pos": [
-              1110,
-              550
-            ],
-            "size": [
-              320,
-              78
-            ],
-            "flags": {},
-            "order": 3,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "link": 13
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "shape": 7,
-                "type": "AUDIO",
-                "link": 16
-              },
-              {
-                "localized_name": "fps",
-                "name": "fps",
-                "type": "FLOAT",
-                "widget": {
-                  "name": "fps"
-                },
-                "link": 12
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "VIDEO",
-                "name": "VIDEO",
-                "type": "VIDEO",
-                "links": [
-                  15
-                ]
-              }
-            ],
-            "properties": {
-              "cnr_id": "comfy-core",
-              "ver": "0.10.0",
-              "Node name for S&R": "CreateVideo"
-            },
-            "widgets_values": [
-              30
-            ]
-          },
-          {
-            "id": 10,
-            "type": "GetVideoComponents",
-            "pos": [
-              1110,
-              330
-            ],
-            "size": [
-              320,
-              70
-            ],
-            "flags": {},
-            "order": 2,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "video",
-                "name": "video",
-                "type": "VIDEO",
-                "link": 10
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "images",
-                "name": "images",
-                "type": "IMAGE",
-                "links": [
-                  14
-                ]
-              },
-              {
-                "localized_name": "audio",
-                "name": "audio",
-                "type": "AUDIO",
-                "links": [
-                  16
-                ]
-              },
-              {
-                "localized_name": "fps",
-                "name": "fps",
-                "type": "FLOAT",
-                "links": [
-                  12
-                ]
-              }
-            ],
-            "properties": {
-              "cnr_id": "comfy-core",
-              "ver": "0.10.0",
-              "Node name for S&R": "GetVideoComponents"
-            }
-          },
-          {
-            "id": 1,
-            "type": "UpscaleModelLoader",
-            "pos": [
-              750,
-              450
-            ],
-            "size": [
-              280,
-              60
-            ],
-            "flags": {},
-            "order": 0,
-            "mode": 0,
-            "inputs": [
-              {
-                "localized_name": "model_name",
-                "name": "model_name",
-                "type": "COMBO",
-                "widget": {
-                  "name": "model_name"
-                },
-                "link": 19
-              }
-            ],
-            "outputs": [
-              {
-                "localized_name": "UPSCALE_MODEL",
-                "name": "UPSCALE_MODEL",
-                "type": "UPSCALE_MODEL",
-                "links": [
-                  1
-                ]
-              }
-            ],
-            "properties": {
-              "cnr_id": "comfy-core",
-              "ver": "0.10.0",
-              "Node name for S&R": "UpscaleModelLoader",
-              "models": [
-                {
-                  "name": "RealESRGAN_x4plus.safetensors",
-                  "url": "https://huggingface.co/Comfy-Org/Real-ESRGAN_repackaged/resolve/main/RealESRGAN_x4plus.safetensors",
-                  "directory": "upscale_models"
-                }
-              ]
-            },
-            "widgets_values": [
-              "RealESRGAN_x4plus.safetensors"
-            ]
-          }
-        ],
-        "groups": [],
-        "links": [
-          {
-            "id": 1,
-            "origin_id": 1,
-            "origin_slot": 0,
-            "target_id": 2,
-            "target_slot": 0,
-            "type": "UPSCALE_MODEL"
-          },
-          {
-            "id": 14,
-            "origin_id": 10,
-            "origin_slot": 0,
-            "target_id": 2,
-            "target_slot": 1,
-            "type": "IMAGE"
-          },
-          {
-            "id": 13,
-            "origin_id": 2,
-            "origin_slot": 0,
-            "target_id": 11,
-            "target_slot": 0,
-            "type": "IMAGE"
-          },
-          {
-            "id": 16,
-            "origin_id": 10,
-            "origin_slot": 1,
-            "target_id": 11,
-            "target_slot": 1,
-            "type": "AUDIO"
-          },
-          {
-            "id": 12,
-            "origin_id": 10,
-            "origin_slot": 2,
-            "target_id": 11,
-            "target_slot": 2,
-            "type": "FLOAT"
-          },
-          {
-            "id": 10,
-            "origin_id": -10,
-            "origin_slot": 0,
-            "target_id": 10,
-            "target_slot": 0,
-            "type": "VIDEO"
-          },
-          {
-            "id": 15,
-            "origin_id": 11,
-            "origin_slot": 0,
-            "target_id": -20,
-            "target_slot": 0,
-            "type": "VIDEO"
-          },
-          {
-            "id": 19,
-            "origin_id": -10,
-            "origin_slot": 1,
-            "target_id": 1,
-            "target_slot": 0,
-            "type": "COMBO"
-          }
-        ],
-        "extra": {
-          "workflowRendererVersion": "LG"
-        },
-        "category": "Video generation and editing/Enhance video",
-        "description": "Upscales video to 4× resolution using a GAN-based upscaling model."
-      }
-    ]
-  },
-  "extra": {}
-}
+{"revision": 0, "last_node_id": 13, "last_link_id": 0, "nodes": [{"id": 13, "type": "cf95b747-3e17-46cb-8097-cac60ff9b2e1", "pos": [1120, 330], "size": [240, 58], "flags": {}, "order": 3, "mode": 0, "inputs": [{"localized_name": "video", "name": "video", "type": "VIDEO", "link": null}, {"name": "model_name", "type": "COMBO", "widget": {"name": "model_name"}, "link": null}], "outputs": [{"localized_name": "VIDEO", "name": "VIDEO", "type": "VIDEO", "links": []}], "title": "Video Upscale(GAN x4)", "properties": {"proxyWidgets": [["-1", "model_name"]], "cnr_id": "comfy-core", "ver": "0.14.1"}, "widgets_values": ["RealESRGAN_x4plus.safetensors"]}], "links": [], "version": 0.4, "definitions": {"subgraphs": [{"id": "cf95b747-3e17-46cb-8097-cac60ff9b2e1", "version": 1, "state": {"lastGroupId": 0, "lastNodeId": 13, "lastLinkId": 19, "lastRerouteId": 0}, "revision": 0, "config": {}, "name": "Video Upscale(GAN x4)", "inputNode": {"id": -10, "bounding": [550, 460, 120, 80]}, "outputNode": {"id": -20, "bounding": [1490, 460, 120, 60]}, "inputs": [{"id": "666d633e-93e7-42dc-8d11-2b7b99b0f2a6", "name": "video", "type": "VIDEO", "linkIds": [10], "localized_name": "video", "pos": [650, 480]}, {"id": "2e23a087-caa8-4d65-99e6-662761aa905a", "name": "model_name", "type": "COMBO", "linkIds": [19], "pos": [650, 500]}], "outputs": [{"id": "0c1768ea-3ec2-412f-9af6-8e0fa36dae70", "name": "VIDEO", "type": "VIDEO", "linkIds": [15], "localized_name": "VIDEO", "pos": [1510, 480]}], "widgets": [], "nodes": [{"id": 2, "type": "ImageUpscaleWithModel", "pos": [1110, 450], "size": [320, 46], "flags": {}, "order": 1, "mode": 0, "inputs": [{"localized_name": "upscale_model", "name": "upscale_model", "type": "UPSCALE_MODEL", "link": 1}, {"localized_name": "image", "name": "image", "type": "IMAGE", "link": 14}], "outputs": [{"localized_name": "IMAGE", "name": "IMAGE", "type": "IMAGE", "links": [13]}], "properties": {"cnr_id": "comfy-core", "ver": "0.10.0", "Node name for S&R": "ImageUpscaleWithModel"}}, {"id": 11, "type": "CreateVideo", "pos": [1110, 550], "size": [320, 78], "flags": {}, "order": 3, "mode": 0, "inputs": [{"localized_name": "images", "name": "images", "type": "IMAGE", "link": 13}, {"localized_name": "audio", "name": "audio", "shape": 7, "type": "AUDIO", "link": 16}, {"localized_name": "fps", "name": "fps", "type": "FLOAT", "widget": {"name": "fps"}, "link": 12}], "outputs": [{"localized_name": "VIDEO", "name": "VIDEO", "type": "VIDEO", "links": [15]}], "properties": {"cnr_id": "comfy-core", "ver": "0.10.0", "Node name for S&R": "CreateVideo"}, "widgets_values": [30]}, {"id": 10, "type": "GetVideoComponents", "pos": [1110, 330], "size": [320, 70], "flags": {}, "order": 2, "mode": 0, "inputs": [{"localized_name": "video", "name": "video", "type": "VIDEO", "link": 10}], "outputs": [{"localized_name": "images", "name": "images", "type": "IMAGE", "links": [14]}, {"localized_name": "audio", "name": "audio", "type": "AUDIO", "links": [16]}, {"localized_name": "fps", "name": "fps", "type": "FLOAT", "links": [12]}], "properties": {"cnr_id": "comfy-core", "ver": "0.10.0", "Node name for S&R": "GetVideoComponents"}}, {"id": 1, "type": "UpscaleModelLoader", "pos": [750, 450], "size": [280, 60], "flags": {}, "order": 0, "mode": 0, "inputs": [{"localized_name": "model_name", "name": "model_name", "type": "COMBO", "widget": {"name": "model_name"}, "link": 19}], "outputs": [{"localized_name": "UPSCALE_MODEL", "name": "UPSCALE_MODEL", "type": "UPSCALE_MODEL", "links": [1]}], "properties": {"cnr_id": "comfy-core", "ver": "0.10.0", "Node name for S&R": "UpscaleModelLoader", "models": [{"name": "RealESRGAN_x4plus.safetensors", "url": "https://huggingface.co/Comfy-Org/Real-ESRGAN_repackaged/resolve/main/RealESRGAN_x4plus.safetensors", "directory": "upscale_models"}]}, "widgets_values": ["RealESRGAN_x4plus.safetensors"]}], "groups": [], "links": [{"id": 1, "origin_id": 1, "origin_slot": 0, "target_id": 2, "target_slot": 0, "type": "UPSCALE_MODEL"}, {"id": 14, "origin_id": 10, "origin_slot": 0, "target_id": 2, "target_slot": 1, "type": "IMAGE"}, {"id": 13, "origin_id": 2, "origin_slot": 0, "target_id": 11, "target_slot": 0, "type": "IMAGE"}, {"id": 16, "origin_id": 10, "origin_slot": 1, "target_id": 11, "target_slot": 1, "type": "AUDIO"}, {"id": 12, "origin_id": 10, "origin_slot": 2, "target_id": 11, "target_slot": 2, "type": "FLOAT"}, {"id": 10, "origin_id": -10, "origin_slot": 0, "target_id": 10, "target_slot": 0, "type": "VIDEO"}, {"id": 15, "origin_id": 11, "origin_slot": 0, "target_id": -20, "target_slot": 0, "type": "VIDEO"}, {"id": 19, "origin_id": -10, "origin_slot": 1, "target_id": 1, "target_slot": 0, "type": "COMBO"}], "extra": {"workflowRendererVersion": "LG"}, "category": "Video generation and editing/Enhance video"}]}, "extra": {}}

comfy/background_removal/birefnet.json

@@ -1,7 +0,0 @@
-{
-    "model_type": "birefnet",
-    "image_std": [1.0, 1.0, 1.0],
-    "image_mean": [0.0, 0.0, 0.0],
-    "image_size": 1024,
-    "resize_to_original": true
-}

comfy/background_removal/birefnet.py

@@ -1,689 +0,0 @@
-import torch
-import comfy.ops
-import numpy as np
-import torch.nn as nn
-from functools import partial
-import torch.nn.functional as F
-from torchvision.ops import deform_conv2d
-from comfy.ldm.modules.attention import optimized_attention_for_device
-
-CXT = [3072, 1536, 768, 384][1:][::-1][-3:]
-
-class Attention(nn.Module):
-    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, device=None, dtype=None, operations=None):
-        super().__init__()
-
-        self.dim = dim
-        self.num_heads = num_heads
-        head_dim = dim // num_heads
-        self.scale = qk_scale or head_dim ** -0.5
-
-        self.q = operations.Linear(dim, dim, bias=qkv_bias, device=device, dtype=dtype)
-        self.kv = operations.Linear(dim, dim * 2, bias=qkv_bias, device=device, dtype=dtype)
-        self.proj = operations.Linear(dim, dim, device=device, dtype=dtype)
-
-    def forward(self, x):
-        B, N, C = x.shape
-        optimized_attention = optimized_attention_for_device(x.device, mask=False, small_input=True)
-        q = self.q(x).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)
-        kv = self.kv(x).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
-        k, v = kv[0], kv[1]
-
-        x = optimized_attention(
-            q, k, v, heads=self.num_heads, skip_output_reshape=True, skip_reshape=True
-        ).transpose(1, 2).reshape(B, N, C)
-        x = self.proj(x)
-
-        return x
-
-class Mlp(nn.Module):
-    def __init__(self, in_features, hidden_features=None, out_features=None, device=None, dtype=None, operations=None):
-        super().__init__()
-        out_features = out_features or in_features
-        hidden_features = hidden_features or in_features
-        self.fc1 = operations.Linear(in_features, hidden_features, device=device, dtype=dtype)
-        self.act = nn.GELU()
-        self.fc2 = operations.Linear(hidden_features, out_features, device=device, dtype=dtype)
-
-    def forward(self, x):
-        x = self.fc1(x)
-        x = self.act(x)
-        x = self.fc2(x)
-        return x
-
-
-def window_partition(x, window_size):
-    B, H, W, C = x.shape
-    x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
-    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
-    return windows
-
-
-def window_reverse(windows, window_size, H, W):
-    B = int(windows.shape[0] / (H * W / window_size / window_size))
-    x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
-    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
-    return x
-
-
-class WindowAttention(nn.Module):
-    def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, device=None, dtype=None, operations=None):
-
-        super().__init__()
-        self.dim = dim
-        self.window_size = window_size  # Wh, Ww
-        self.num_heads = num_heads
-        head_dim = dim // num_heads
-        self.scale = qk_scale or head_dim ** -0.5
-
-        self.relative_position_bias_table = nn.Parameter(
-            torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads, device=device, dtype=dtype))
-
-        coords_h = torch.arange(self.window_size[0])
-        coords_w = torch.arange(self.window_size[1])
-        coords = torch.stack(torch.meshgrid([coords_h, coords_w], indexing='ij'))  # 2, Wh, Ww
-        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
-        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
-        relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
-        relative_coords[:, :, 0] += self.window_size[0] - 1
-        relative_coords[:, :, 1] += self.window_size[1] - 1
-        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
-        relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
-        self.register_buffer("relative_position_index", relative_position_index)
-
-        self.qkv = operations.Linear(dim, dim * 3, bias=qkv_bias, device=device, dtype=dtype)
-        self.proj = operations.Linear(dim, dim, device=device, dtype=dtype)
-        self.softmax = nn.Softmax(dim=-1)
-
-    def forward(self, x, mask=None):
-        B_, N, C = x.shape
-        qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
-        q, k, v = qkv[0], qkv[1], qkv[2]
-
-        q = q * self.scale
-        attn = (q @ k.transpose(-2, -1))
-
-        relative_position_bias = self.relative_position_bias_table[self.relative_position_index.long().view(-1)].view(
-            self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1)  # Wh*Ww,Wh*Ww,nH
-        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
-        attn = attn + relative_position_bias.unsqueeze(0)
-
-        if mask is not None:
-            nW = mask.shape[0]
-            attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
-            attn = attn.view(-1, self.num_heads, N, N)
-            attn = self.softmax(attn)
-        else:
-            attn = self.softmax(attn)
-
-        x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
-        x = self.proj(x)
-        return x
-
-
-class SwinTransformerBlock(nn.Module):
-    def __init__(self, dim, num_heads, window_size=7, shift_size=0,
-                 mlp_ratio=4., qkv_bias=True, qk_scale=None,
-                 norm_layer=nn.LayerNorm, device=None, dtype=None, operations=None):
-        super().__init__()
-        self.dim = dim
-        self.num_heads = num_heads
-        self.window_size = window_size
-        self.shift_size = shift_size
-        self.mlp_ratio = mlp_ratio
-
-        self.norm1 = norm_layer(dim, device=device, dtype=dtype)
-        self.attn = WindowAttention(
-            dim, window_size=(self.window_size, self.window_size), num_heads=num_heads,
-            qkv_bias=qkv_bias, qk_scale=qk_scale, device=device, dtype=dtype, operations=operations)
-
-        self.norm2 = norm_layer(dim, device=device, dtype=dtype)
-        mlp_hidden_dim = int(dim * mlp_ratio)
-        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, device=device, dtype=dtype, operations=operations)
-
-        self.H = None
-        self.W = None
-
-    def forward(self, x, mask_matrix):
-        B, L, C = x.shape
-        H, W = self.H, self.W
-
-        shortcut = x
-        x = self.norm1(x)
-        x = x.view(B, H, W, C)
-
-        pad_l = pad_t = 0
-        pad_r = (self.window_size - W % self.window_size) % self.window_size
-        pad_b = (self.window_size - H % self.window_size) % self.window_size
-        x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b))
-        _, Hp, Wp, _ = x.shape
-
-        if self.shift_size > 0:
-            shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
-            attn_mask = mask_matrix
-        else:
-            shifted_x = x
-            attn_mask = None
-
-        x_windows = window_partition(shifted_x, self.window_size)
-        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)
-
-        attn_windows = self.attn(x_windows, mask=attn_mask)
-
-        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)
-        shifted_x = window_reverse(attn_windows, self.window_size, Hp, Wp)  # B H' W' C
-
-        if self.shift_size > 0:
-            x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
-        else:
-            x = shifted_x
-
-        if pad_r > 0 or pad_b > 0:
-            x = x[:, :H, :W, :].contiguous()
-
-        x = x.view(B, H * W, C)
-
-        x = shortcut + x
-        x = x + self.mlp(self.norm2(x))
-
-        return x
-
-
-class PatchMerging(nn.Module):
-    def __init__(self, dim, device=None, dtype=None, operations=None):
-        super().__init__()
-        self.dim = dim
-        self.reduction = operations.Linear(4 * dim, 2 * dim, bias=False, device=device, dtype=dtype)
-        self.norm = operations.LayerNorm(4 * dim, device=device, dtype=dtype)
-
-    def forward(self, x, H, W):
-        B, L, C = x.shape
-        x = x.view(B, H, W, C)
-
-        # padding
-        pad_input = (H % 2 == 1) or (W % 2 == 1)
-        if pad_input:
-            x = F.pad(x, (0, 0, 0, W % 2, 0, H % 2))
-
-        x0 = x[:, 0::2, 0::2, :]  # B H/2 W/2 C
-        x1 = x[:, 1::2, 0::2, :]  # B H/2 W/2 C
-        x2 = x[:, 0::2, 1::2, :]  # B H/2 W/2 C
-        x3 = x[:, 1::2, 1::2, :]  # B H/2 W/2 C
-        x = torch.cat([x0, x1, x2, x3], -1)  # B H/2 W/2 4*C
-        x = x.view(B, -1, 4 * C)  # B H/2*W/2 4*C
-
-        x = self.norm(x)
-        x = self.reduction(x)
-
-        return x
-
-
-class BasicLayer(nn.Module):
-    def __init__(self,
-                 dim,
-                 depth,
-                 num_heads,
-                 window_size=7,
-                 mlp_ratio=4.,
-                 qkv_bias=True,
-                 qk_scale=None,
-                 norm_layer=nn.LayerNorm,
-                 downsample=None,
-                 device=None, dtype=None, operations=None):
-        super().__init__()
-        self.window_size = window_size
-        self.shift_size = window_size // 2
-        self.depth = depth
-
-        # build blocks
-        self.blocks = nn.ModuleList([
-            SwinTransformerBlock(
-                dim=dim,
-                num_heads=num_heads,
-                window_size=window_size,
-                shift_size=0 if (i % 2 == 0) else window_size // 2,
-                mlp_ratio=mlp_ratio,
-                qkv_bias=qkv_bias,
-                qk_scale=qk_scale,
-                norm_layer=norm_layer,
-                device=device, dtype=dtype, operations=operations)
-            for i in range(depth)])
-
-        # patch merging layer
-        if downsample is not None:
-            self.downsample = downsample(dim=dim, device=device, dtype=dtype, operations=operations)
-        else:
-            self.downsample = None
-
-    def forward(self, x, H, W):
-        Hp = int(np.ceil(H / self.window_size)) * self.window_size
-        Wp = int(np.ceil(W / self.window_size)) * self.window_size
-        img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device)  # 1 Hp Wp 1
-        h_slices = (slice(0, -self.window_size),
-                    slice(-self.window_size, -self.shift_size),
-                    slice(-self.shift_size, None))
-        w_slices = (slice(0, -self.window_size),
-                    slice(-self.window_size, -self.shift_size),
-                    slice(-self.shift_size, None))
-        cnt = 0
-        for h in h_slices:
-            for w in w_slices:
-                img_mask[:, h, w, :] = cnt
-                cnt += 1
-
-        mask_windows = window_partition(img_mask, self.window_size)
-        mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
-        attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
-        attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))
-
-        for blk in self.blocks:
-            blk.H, blk.W = H, W
-            x = blk(x, attn_mask)
-        if self.downsample is not None:
-            x_down = self.downsample(x, H, W)
-            Wh, Ww = (H + 1) // 2, (W + 1) // 2
-            return x, H, W, x_down, Wh, Ww
-        else:
-            return x, H, W, x, H, W
-
-
-class PatchEmbed(nn.Module):
-    def __init__(self, patch_size=4, in_channels=3, embed_dim=96, norm_layer=None, device=None, dtype=None, operations=None):
-        super().__init__()
-        patch_size = (patch_size, patch_size)
-        self.patch_size = patch_size
-
-        self.in_channels = in_channels
-        self.embed_dim = embed_dim
-
-        self.proj = operations.Conv2d(in_channels, embed_dim, kernel_size=patch_size, stride=patch_size, device=device, dtype=dtype)
-        if norm_layer is not None:
-            self.norm = norm_layer(embed_dim, device=device, dtype=dtype)
-        else:
-            self.norm = None
-
-    def forward(self, x):
-        _, _, H, W = x.size()
-        if W % self.patch_size[1] != 0:
-            x = F.pad(x, (0, self.patch_size[1] - W % self.patch_size[1]))
-        if H % self.patch_size[0] != 0:
-            x = F.pad(x, (0, 0, 0, self.patch_size[0] - H % self.patch_size[0]))
-
-        x = self.proj(x)  # B C Wh Ww
-        if self.norm is not None:
-            Wh, Ww = x.size(2), x.size(3)
-            x = x.flatten(2).transpose(1, 2)
-            x = self.norm(x)
-            x = x.transpose(1, 2).view(-1, self.embed_dim, Wh, Ww)
-
-        return x
-
-
-class SwinTransformer(nn.Module):
-    def __init__(self,
-                 pretrain_img_size=224,
-                 patch_size=4,
-                 in_channels=3,
-                 embed_dim=96,
-                 depths=[2, 2, 6, 2],
-                 num_heads=[3, 6, 12, 24],
-                 window_size=7,
-                 mlp_ratio=4.,
-                 qkv_bias=True,
-                 qk_scale=None,
-                 patch_norm=True,
-                 out_indices=(0, 1, 2, 3),
-                 frozen_stages=-1,
-                 device=None, dtype=None, operations=None):
-        super().__init__()
-
-        norm_layer = partial(operations.LayerNorm, device=device, dtype=dtype)
-        self.pretrain_img_size = pretrain_img_size
-        self.num_layers = len(depths)
-        self.embed_dim = embed_dim
-        self.patch_norm = patch_norm
-        self.out_indices = out_indices
-        self.frozen_stages = frozen_stages
-
-        self.patch_embed = PatchEmbed(
-            patch_size=patch_size, in_channels=in_channels, embed_dim=embed_dim,
-            device=device, dtype=dtype, operations=operations,
-            norm_layer=norm_layer if self.patch_norm else None)
-
-        self.layers = nn.ModuleList()
-        for i_layer in range(self.num_layers):
-            layer = BasicLayer(
-                dim=int(embed_dim * 2 ** i_layer),
-                depth=depths[i_layer],
-                num_heads=num_heads[i_layer],
-                window_size=window_size,
-                mlp_ratio=mlp_ratio,
-                qkv_bias=qkv_bias,
-                qk_scale=qk_scale,
-                norm_layer=norm_layer,
-                downsample=PatchMerging if (i_layer < self.num_layers - 1) else None,
-                device=device, dtype=dtype, operations=operations)
-            self.layers.append(layer)
-
-        num_features = [int(embed_dim * 2 ** i) for i in range(self.num_layers)]
-        self.num_features = num_features
-
-        for i_layer in out_indices:
-            layer = norm_layer(num_features[i_layer])
-            layer_name = f'norm{i_layer}'
-            self.add_module(layer_name, layer)
-
-
-    def forward(self, x):
-        x = self.patch_embed(x)
-
-        Wh, Ww = x.size(2), x.size(3)
-
-        outs = []
-        x = x.flatten(2).transpose(1, 2)
-        for i in range(self.num_layers):
-            layer = self.layers[i]
-            x_out, H, W, x, Wh, Ww = layer(x, Wh, Ww)
-
-            if i in self.out_indices:
-                norm_layer = getattr(self, f'norm{i}')
-                x_out = norm_layer(x_out)
-
-                out = x_out.view(-1, H, W, self.num_features[i]).permute(0, 3, 1, 2).contiguous()
-                outs.append(out)
-
-        return tuple(outs)
-
-class DeformableConv2d(nn.Module):
-    def __init__(self,
-                 in_channels,
-                 out_channels,
-                 kernel_size=3,
-                 stride=1,
-                 padding=1,
-                 bias=False, device=None, dtype=None, operations=None):
-
-        super(DeformableConv2d, self).__init__()
-
-        kernel_size = kernel_size if type(kernel_size) is tuple else (kernel_size, kernel_size)
-        self.stride = stride if type(stride) is tuple else (stride, stride)
-        self.padding = padding
-
-        self.offset_conv = operations.Conv2d(in_channels,
-                                     2 * kernel_size[0] * kernel_size[1],
-                                     kernel_size=kernel_size,
-                                     stride=stride,
-                                     padding=self.padding,
-                                     bias=True, device=device, dtype=dtype)
-
-        self.modulator_conv = operations.Conv2d(in_channels,
-                                     1 * kernel_size[0] * kernel_size[1],
-                                     kernel_size=kernel_size,
-                                     stride=stride,
-                                     padding=self.padding,
-                                     bias=True, device=device, dtype=dtype)
-
-        self.regular_conv = operations.Conv2d(in_channels,
-                                      out_channels=out_channels,
-                                      kernel_size=kernel_size,
-                                      stride=stride,
-                                      padding=self.padding,
-                                      bias=bias, device=device, dtype=dtype)
-
-    def forward(self, x):
-        offset = self.offset_conv(x)
-        modulator = 2. * torch.sigmoid(self.modulator_conv(x))
-        weight, bias, offload_info = comfy.ops.cast_bias_weight(self.regular_conv, x, offloadable=True)
-
-        x = deform_conv2d(
-            input=x,
-            offset=offset,
-            weight=weight,
-            bias=None,
-            padding=self.padding,
-            mask=modulator,
-            stride=self.stride,
-        )
-        comfy.ops.uncast_bias_weight(self.regular_conv, weight, bias, offload_info)
-        return x
-
-class BasicDecBlk(nn.Module):
-    def __init__(self, in_channels=64, out_channels=64, inter_channels=64, device=None, dtype=None, operations=None):
-        super(BasicDecBlk, self).__init__()
-        inter_channels = 64
-        self.conv_in = operations.Conv2d(in_channels, inter_channels, 3, 1, padding=1, device=device, dtype=dtype)
-        self.relu_in = nn.ReLU(inplace=True)
-        self.dec_att = ASPPDeformable(in_channels=inter_channels, device=device, dtype=dtype, operations=operations)
-        self.conv_out = operations.Conv2d(inter_channels, out_channels, 3, 1, padding=1, device=device, dtype=dtype)
-        self.bn_in = operations.BatchNorm2d(inter_channels, device=device, dtype=dtype)
-        self.bn_out = operations.BatchNorm2d(out_channels, device=device, dtype=dtype)
-
-    def forward(self, x):
-        x = self.conv_in(x)
-        x = self.bn_in(x)
-        x = self.relu_in(x)
-        x = self.dec_att(x)
-        x = self.conv_out(x)
-        x = self.bn_out(x)
-        return x
-
-
-class BasicLatBlk(nn.Module):
-    def __init__(self, in_channels=64, out_channels=64, device=None, dtype=None, operations=None):
-        super(BasicLatBlk, self).__init__()
-        self.conv = operations.Conv2d(in_channels, out_channels, 1, 1, 0, device=device, dtype=dtype)
-
-    def forward(self, x):
-        x = self.conv(x)
-        return x
-
-
-class _ASPPModuleDeformable(nn.Module):
-    def __init__(self, in_channels, planes, kernel_size, padding, device, dtype, operations):
-        super(_ASPPModuleDeformable, self).__init__()
-        self.atrous_conv = DeformableConv2d(in_channels, planes, kernel_size=kernel_size,
-                                            stride=1, padding=padding, bias=False, device=device, dtype=dtype, operations=operations)
-        self.bn = operations.BatchNorm2d(planes, device=device, dtype=dtype)
-        self.relu = nn.ReLU(inplace=True)
-
-    def forward(self, x):
-        x = self.atrous_conv(x)
-        x = self.bn(x)
-
-        return self.relu(x)
-
-
-class ASPPDeformable(nn.Module):
-    def __init__(self, in_channels, out_channels=None, parallel_block_sizes=[1, 3, 7], device=None, dtype=None, operations=None):
-        super(ASPPDeformable, self).__init__()
-        self.down_scale = 1
-        if out_channels is None:
-            out_channels = in_channels
-        self.in_channelster = 256 // self.down_scale
-
-        self.aspp1 = _ASPPModuleDeformable(in_channels, self.in_channelster, 1, padding=0, device=device, dtype=dtype, operations=operations)
-        self.aspp_deforms = nn.ModuleList([
-            _ASPPModuleDeformable(in_channels, self.in_channelster, conv_size, padding=int(conv_size//2), device=device, dtype=dtype, operations=operations)
-              for conv_size in parallel_block_sizes
-        ])
-
-        self.global_avg_pool = nn.Sequential(nn.AdaptiveAvgPool2d((1, 1)),
-                                             operations.Conv2d(in_channels, self.in_channelster, 1, stride=1, bias=False, device=device, dtype=dtype),
-                                             operations.BatchNorm2d(self.in_channelster, device=device, dtype=dtype),
-                                             nn.ReLU(inplace=True))
-        self.conv1 = operations.Conv2d(self.in_channelster * (2 + len(self.aspp_deforms)), out_channels, 1, bias=False, device=device, dtype=dtype)
-        self.bn1 = operations.BatchNorm2d(out_channels, device=device, dtype=dtype)
-        self.relu = nn.ReLU(inplace=True)
-
-    def forward(self, x):
-        x1 = self.aspp1(x)
-        x_aspp_deforms = [aspp_deform(x) for aspp_deform in self.aspp_deforms]
-        x5 = self.global_avg_pool(x)
-        x5 = F.interpolate(x5, size=x1.size()[2:], mode='bilinear', align_corners=True)
-        x = torch.cat((x1, *x_aspp_deforms, x5), dim=1)
-
-        x = self.conv1(x)
-        x = self.bn1(x)
-        x = self.relu(x)
-
-        return x
-
-class BiRefNet(nn.Module):
-    def __init__(self, config=None, dtype=None, device=None, operations=None):
-        super(BiRefNet, self).__init__()
-        self.bb = SwinTransformer(embed_dim=192, depths=[2, 2, 18, 2], num_heads=[6, 12, 24, 48], window_size=12, device=device, dtype=dtype, operations=operations)
-
-        channels = [1536, 768, 384, 192]
-        channels = [c * 2 for c in channels]
-        self.cxt = channels[1:][::-1][-3:]
-        self.squeeze_module = nn.Sequential(*[
-            BasicDecBlk(channels[0]+sum(self.cxt), channels[0], device=device, dtype=dtype, operations=operations)
-            for _ in range(1)
-        ])
-
-        self.decoder = Decoder(channels, device=device, dtype=dtype, operations=operations)
-
-    def forward_enc(self, x):
-        x1, x2, x3, x4 = self.bb(x)
-        B, C, H, W = x.shape
-        x1_, x2_, x3_, x4_ = self.bb(F.interpolate(x, size=(H//2, W//2), mode='bilinear', align_corners=True))
-        x1 = torch.cat([x1, F.interpolate(x1_, size=x1.shape[2:], mode='bilinear', align_corners=True)], dim=1)
-        x2 = torch.cat([x2, F.interpolate(x2_, size=x2.shape[2:], mode='bilinear', align_corners=True)], dim=1)
-        x3 = torch.cat([x3, F.interpolate(x3_, size=x3.shape[2:], mode='bilinear', align_corners=True)], dim=1)
-        x4 = torch.cat([x4, F.interpolate(x4_, size=x4.shape[2:], mode='bilinear', align_corners=True)], dim=1)
-        x4 = torch.cat(
-            (
-                *[
-                    F.interpolate(x1, size=x4.shape[2:], mode='bilinear', align_corners=True),
-                    F.interpolate(x2, size=x4.shape[2:], mode='bilinear', align_corners=True),
-                    F.interpolate(x3, size=x4.shape[2:], mode='bilinear', align_corners=True),
-                ][-len(CXT):],
-                x4
-            ),
-            dim=1
-        )
-        return (x1, x2, x3, x4)
-
-    def forward_ori(self, x):
-        (x1, x2, x3, x4) = self.forward_enc(x)
-        x4 = self.squeeze_module(x4)
-        features = [x, x1, x2, x3, x4]
-        scaled_preds = self.decoder(features)
-        return scaled_preds
-
-    def forward(self, pixel_values, intermediate_output=None):
-        scaled_preds = self.forward_ori(pixel_values)
-        return scaled_preds
-
-
-class Decoder(nn.Module):
-    def __init__(self, channels, device, dtype, operations):
-        super(Decoder, self).__init__()
-        # factory kwargs
-        fk = {"device":device, "dtype":dtype, "operations":operations}
-        DecoderBlock = partial(BasicDecBlk, **fk)
-        LateralBlock = partial(BasicLatBlk, **fk)
-        DBlock = partial(SimpleConvs, **fk)
-
-        self.split = True
-        N_dec_ipt = 64
-        ic = 64
-        ipt_cha_opt = 1
-        self.ipt_blk5 = DBlock(2**10*3 if self.split else 3, [N_dec_ipt, channels[0]//8][ipt_cha_opt], inter_channels=ic)
-        self.ipt_blk4 = DBlock(2**8*3 if self.split else 3, [N_dec_ipt, channels[0]//8][ipt_cha_opt], inter_channels=ic)
-        self.ipt_blk3 = DBlock(2**6*3 if self.split else 3, [N_dec_ipt, channels[1]//8][ipt_cha_opt], inter_channels=ic)
-        self.ipt_blk2 = DBlock(2**4*3 if self.split else 3, [N_dec_ipt, channels[2]//8][ipt_cha_opt], inter_channels=ic)
-        self.ipt_blk1 = DBlock(2**0*3 if self.split else 3, [N_dec_ipt, channels[3]//8][ipt_cha_opt], inter_channels=ic)
-
-        self.decoder_block4 = DecoderBlock(channels[0]+([N_dec_ipt, channels[0]//8][ipt_cha_opt]), channels[1])
-        self.decoder_block3 = DecoderBlock(channels[1]+([N_dec_ipt, channels[0]//8][ipt_cha_opt]), channels[2])
-        self.decoder_block2 = DecoderBlock(channels[2]+([N_dec_ipt, channels[1]//8][ipt_cha_opt]), channels[3])
-        self.decoder_block1 = DecoderBlock(channels[3]+([N_dec_ipt, channels[2]//8][ipt_cha_opt]), channels[3]//2)
-
-        fk = {"device":device, "dtype":dtype}
-
-        self.conv_out1 = nn.Sequential(operations.Conv2d(channels[3]//2+([N_dec_ipt, channels[3]//8][ipt_cha_opt]), 1, 1, 1, 0, **fk))
-
-        self.lateral_block4 = LateralBlock(channels[1], channels[1])
-        self.lateral_block3 = LateralBlock(channels[2], channels[2])
-        self.lateral_block2 = LateralBlock(channels[3], channels[3])
-
-        self.conv_ms_spvn_4 = operations.Conv2d(channels[1], 1, 1, 1, 0, **fk)
-        self.conv_ms_spvn_3 = operations.Conv2d(channels[2], 1, 1, 1, 0, **fk)
-        self.conv_ms_spvn_2 = operations.Conv2d(channels[3], 1, 1, 1, 0, **fk)
-
-        _N = 16
-
-        self.gdt_convs_4 = nn.Sequential(operations.Conv2d(channels[0] // 2, _N, 3, 1, 1, **fk), operations.BatchNorm2d(_N, **fk), nn.ReLU(inplace=True))
-        self.gdt_convs_3 = nn.Sequential(operations.Conv2d(channels[1] // 2, _N, 3, 1, 1, **fk), operations.BatchNorm2d(_N, **fk), nn.ReLU(inplace=True))
-        self.gdt_convs_2 = nn.Sequential(operations.Conv2d(channels[2] // 2, _N, 3, 1, 1, **fk), operations.BatchNorm2d(_N, **fk), nn.ReLU(inplace=True))
-
-        [setattr(self, f"gdt_convs_pred_{i}", nn.Sequential(operations.Conv2d(_N, 1, 1, 1, 0, **fk))) for i in range(2, 5)]
-        [setattr(self, f"gdt_convs_attn_{i}", nn.Sequential(operations.Conv2d(_N, 1, 1, 1, 0, **fk))) for i in range(2, 5)]
-
-    def get_patches_batch(self, x, p):
-        _size_h, _size_w = p.shape[2:]
-        patches_batch = []
-        for idx in range(x.shape[0]):
-            columns_x = torch.split(x[idx], split_size_or_sections=_size_w, dim=-1)
-            patches_x = []
-            for column_x in columns_x:
-                patches_x += [p.unsqueeze(0) for p in torch.split(column_x, split_size_or_sections=_size_h, dim=-2)]
-            patch_sample = torch.cat(patches_x, dim=1)
-            patches_batch.append(patch_sample)
-        return torch.cat(patches_batch, dim=0)
-
-    def forward(self, features):
-        x, x1, x2, x3, x4 = features
-
-        patches_batch = self.get_patches_batch(x, x4) if self.split else x
-        x4 = torch.cat((x4, self.ipt_blk5(F.interpolate(patches_batch, size=x4.shape[2:], mode='bilinear', align_corners=True))), 1)
-        p4 = self.decoder_block4(x4)
-        p4_gdt = self.gdt_convs_4(p4)
-        gdt_attn_4 = self.gdt_convs_attn_4(p4_gdt).sigmoid()
-        p4 = p4 * gdt_attn_4
-        _p4 = F.interpolate(p4, size=x3.shape[2:], mode='bilinear', align_corners=True)
-        _p3 = _p4 + self.lateral_block4(x3)
-
-        patches_batch = self.get_patches_batch(x, _p3) if self.split else x
-        _p3 = torch.cat((_p3, self.ipt_blk4(F.interpolate(patches_batch, size=x3.shape[2:], mode='bilinear', align_corners=True))), 1)
-        p3 = self.decoder_block3(_p3)
-
-        p3_gdt = self.gdt_convs_3(p3)
-        gdt_attn_3 = self.gdt_convs_attn_3(p3_gdt).sigmoid()
-        p3 = p3 * gdt_attn_3
-        _p3 = F.interpolate(p3, size=x2.shape[2:], mode='bilinear', align_corners=True)
-        _p2 = _p3 + self.lateral_block3(x2)
-
-        patches_batch = self.get_patches_batch(x, _p2) if self.split else x
-        _p2 = torch.cat((_p2, self.ipt_blk3(F.interpolate(patches_batch, size=x2.shape[2:], mode='bilinear', align_corners=True))), 1)
-        p2 = self.decoder_block2(_p2)
-
-        p2_gdt = self.gdt_convs_2(p2)
-        gdt_attn_2 = self.gdt_convs_attn_2(p2_gdt).sigmoid()
-        p2 = p2 * gdt_attn_2
-
-        _p2 = F.interpolate(p2, size=x1.shape[2:], mode='bilinear', align_corners=True)
-        _p1 = _p2 + self.lateral_block2(x1)
-
-        patches_batch = self.get_patches_batch(x, _p1) if self.split else x
-        _p1 = torch.cat((_p1, self.ipt_blk2(F.interpolate(patches_batch, size=x1.shape[2:], mode='bilinear', align_corners=True))), 1)
-        _p1 = self.decoder_block1(_p1)
-        _p1 = F.interpolate(_p1, size=x.shape[2:], mode='bilinear', align_corners=True)
-
-        patches_batch = self.get_patches_batch(x, _p1) if self.split else x
-        _p1 = torch.cat((_p1, self.ipt_blk1(F.interpolate(patches_batch, size=x.shape[2:], mode='bilinear', align_corners=True))), 1)
-        p1_out = self.conv_out1(_p1)
-        return p1_out
-
-
-class SimpleConvs(nn.Module):
-    def __init__(
-        self, in_channels: int, out_channels: int, inter_channels=64, device=None, dtype=None, operations=None
-    ) -> None:
-        super().__init__()
-        self.conv1 = operations.Conv2d(in_channels, inter_channels, 3, 1, 1, device=device, dtype=dtype)
-        self.conv_out = operations.Conv2d(inter_channels, out_channels, 3, 1, 1, device=device, dtype=dtype)
-
-    def forward(self, x):
-        return self.conv_out(self.conv1(x))

comfy/bg_removal_model.py

@@ -1,78 +0,0 @@
-from .utils import load_torch_file
-import os
-import json
-import torch
-import logging
-
-import comfy.ops
-import comfy.model_patcher
-import comfy.model_management
-import comfy.clip_model
-import comfy.background_removal.birefnet
-
-BG_REMOVAL_MODELS = {
-    "birefnet": comfy.background_removal.birefnet.BiRefNet
-}
-
-class BackgroundRemovalModel():
-    def __init__(self, json_config):
-        with open(json_config) as f:
-            config = json.load(f)
-
-        self.image_size = config.get("image_size", 1024)
-        self.image_mean = config.get("image_mean", [0.0, 0.0, 0.0])
-        self.image_std = config.get("image_std", [1.0, 1.0, 1.0])
-        self.model_type = config.get("model_type", "birefnet")
-        self.config = config.copy()
-        model_class = BG_REMOVAL_MODELS.get(self.model_type)
-
-        self.load_device = comfy.model_management.text_encoder_device()
-        offload_device = comfy.model_management.text_encoder_offload_device()
-        self.dtype = comfy.model_management.text_encoder_dtype(self.load_device)
-        self.model = model_class(config, self.dtype, offload_device, comfy.ops.manual_cast)
-        self.model.eval()
-
-        self.patcher = comfy.model_patcher.CoreModelPatcher(self.model, load_device=self.load_device, offload_device=offload_device)
-
-    def load_sd(self, sd):
-        return self.model.load_state_dict(sd, strict=False, assign=self.patcher.is_dynamic())
-
-    def get_sd(self):
-        return self.model.state_dict()
-
-    def encode_image(self, image):
-        comfy.model_management.load_model_gpu(self.patcher)
-        H, W = image.shape[1], image.shape[2]
-        pixel_values = comfy.clip_model.clip_preprocess(image.to(self.load_device), size=self.image_size, mean=self.image_mean, std=self.image_std, crop=False)
-        out = self.model(pixel_values=pixel_values)
-        out = torch.nn.functional.interpolate(out, size=(H, W), mode="bicubic", antialias=False)
-
-        mask = out.sigmoid().to(device=comfy.model_management.intermediate_device(), dtype=comfy.model_management.intermediate_dtype())
-        if mask.ndim == 3:
-            mask = mask.unsqueeze(0)
-        if mask.shape[1] != 1:
-            mask = mask.movedim(-1, 1)
-
-        return mask
-
-
-def load_background_removal_model(sd):
-    if "bb.layers.1.blocks.0.attn.relative_position_index" in sd:
-        json_config = os.path.join(os.path.join(os.path.dirname(os.path.realpath(__file__)), "background_removal"), "birefnet.json")
-    else:
-        return None
-
-    bg_model = BackgroundRemovalModel(json_config)
-    m, u = bg_model.load_sd(sd)
-    if len(m) > 0:
-        logging.warning("missing background removal: {}".format(m))
-    u = set(u)
-    keys = list(sd.keys())
-    for k in keys:
-        if k not in u:
-            sd.pop(k)
-    return bg_model
-
-def load(ckpt_path):
-    sd = load_torch_file(ckpt_path)
-    return load_background_removal_model(sd)

comfy/cli_args.py

@@ -90,8 +90,8 @@ parser.add_argument("--force-channels-last", action="store_true", help="Force ch
 parser.add_argument("--directml", type=int, nargs="?", metavar="DIRECTML_DEVICE", const=-1, help="Use torch-directml.")
 
 parser.add_argument("--oneapi-device-selector", type=str, default=None, metavar="SELECTOR_STRING", help="Sets the oneAPI device(s) this instance will use.")
+parser.add_argument("--disable-ipex-optimize", action="store_true", help="Disables ipex.optimize default when loading models with Intel's Extension for Pytorch.")
 parser.add_argument("--supports-fp8-compute", action="store_true", help="ComfyUI will act like if the device supports fp8 compute.")
-parser.add_argument("--enable-triton-backend", action="store_true", help="ComfyUI will enable the use of Triton backend in comfy-kitchen. Is disabled at launch by default.")
 
 class LatentPreviewMethod(enum.Enum):
     NoPreviews = "none"
@@ -238,8 +238,6 @@ database_default_path = os.path.abspath(
 )
 parser.add_argument("--database-url", type=str, default=f"sqlite:///{database_default_path}", help="Specify the database URL, e.g. for an in-memory database you can use 'sqlite:///:memory:'.")
 parser.add_argument("--enable-assets", action="store_true", help="Enable the assets system (API routes, database synchronization, and background scanning).")
-parser.add_argument("--feature-flag", type=str, action='append', default=[], metavar="KEY[=VALUE]", help="Set a server feature flag. Use KEY=VALUE to set an explicit value, or bare KEY to set it to true. Can be specified multiple times. Boolean values (true/false) and numbers are auto-converted. Examples: --feature-flag show_signin_button=true  or  --feature-flag show_signin_button")
-parser.add_argument("--list-feature-flags", action="store_true", help="Print the registry of known CLI-settable feature flags as JSON and exit.")
 
 if comfy.options.args_parsing:
     args = parser.parse_args()

comfy/context_windows.py

@@ -63,11 +63,7 @@ class IndexListContextWindow(ContextWindowABC):
             dim = self.dim
         if dim == 0 and full.shape[dim] == 1:
             return full
-        indices = self.index_list
-        anchor_idx = getattr(self, 'causal_anchor_index', None)
-        if anchor_idx is not None and anchor_idx >= 0:
-            indices = [anchor_idx] + list(indices)
-        idx = tuple([slice(None)] * dim + [indices])
+        idx = tuple([slice(None)] * dim + [self.index_list])
         window = full[idx]
         if retain_index_list:
             idx = tuple([slice(None)] * dim + [retain_index_list])
@@ -117,14 +113,7 @@ def slice_cond(cond_value, window: IndexListContextWindow, x_in: torch.Tensor, d
 
     # skip leading latent positions that have no corresponding conditioning (e.g. reference frames)
     if temporal_offset > 0:
-        anchor_idx = getattr(window, 'causal_anchor_index', None)
-        if anchor_idx is not None and anchor_idx >= 0:
-            # anchor occupies one of the no-cond positions, so skip one fewer from window.index_list
-            skip_count = temporal_offset - 1
-        else:
-            skip_count = temporal_offset
-
-        indices = [i - temporal_offset for i in window.index_list[skip_count:]]
+        indices = [i - temporal_offset for i in window.index_list[temporal_offset:]]
         indices = [i for i in indices if 0 <= i]
     else:
         indices = list(window.index_list)
@@ -161,8 +150,7 @@ class ContextFuseMethod:
 ContextResults = collections.namedtuple("ContextResults", ['window_idx', 'sub_conds_out', 'sub_conds', 'window'])
 class IndexListContextHandler(ContextHandlerABC):
     def __init__(self, context_schedule: ContextSchedule, fuse_method: ContextFuseMethod, context_length: int=1, context_overlap: int=0, context_stride: int=1,
-                 closed_loop: bool=False, dim:int=0, freenoise: bool=False, cond_retain_index_list: list[int]=[], split_conds_to_windows: bool=False,
-                 causal_window_fix: bool=True):
+                 closed_loop: bool=False, dim:int=0, freenoise: bool=False, cond_retain_index_list: list[int]=[], split_conds_to_windows: bool=False):
         self.context_schedule = context_schedule
         self.fuse_method = fuse_method
         self.context_length = context_length
@@ -174,7 +162,6 @@ class IndexListContextHandler(ContextHandlerABC):
         self.freenoise = freenoise
         self.cond_retain_index_list = [int(x.strip()) for x in cond_retain_index_list.split(",")] if cond_retain_index_list else []
         self.split_conds_to_windows = split_conds_to_windows
-        self.causal_window_fix = causal_window_fix
 
         self.callbacks = {}
 
@@ -331,14 +318,6 @@ class IndexListContextHandler(ContextHandlerABC):
             # allow processing to end between context window executions for faster Cancel
             comfy.model_management.throw_exception_if_processing_interrupted()
 
-            # causal_window_fix: prepend a pre-window frame that will be stripped post-forward
-            anchor_applied = False
-            if self.causal_window_fix:
-                anchor_idx = window.index_list[0] - 1
-                if 0 <= anchor_idx < x_in.size(self.dim):
-                    window.causal_anchor_index = anchor_idx
-                    anchor_applied = True
-
             for callback in comfy.patcher_extension.get_all_callbacks(IndexListCallbacks.EVALUATE_CONTEXT_WINDOWS, self.callbacks):
                 callback(self, model, x_in, conds, timestep, model_options, window_idx, window, model_options, device, first_device)
 
@@ -353,12 +332,6 @@ class IndexListContextHandler(ContextHandlerABC):
             if device is not None:
                 for i in range(len(sub_conds_out)):
                     sub_conds_out[i] = sub_conds_out[i].to(x_in.device)
-
-            # strip causal_window_fix anchor if applied
-            if anchor_applied:
-                for i in range(len(sub_conds_out)):
-                    sub_conds_out[i] = sub_conds_out[i].narrow(self.dim, 1, sub_conds_out[i].shape[self.dim] - 1)
-
             results.append(ContextResults(window_idx, sub_conds_out, sub_conds, window))
         return results
 

comfy/deploy_environment.py

@@ -1,34 +0,0 @@
-import functools
-import logging
-import os
-
-logger = logging.getLogger(__name__)
-
-_DEFAULT_DEPLOY_ENV = "local-git"
-_ENV_FILENAME = ".comfy_environment"
-
-# Resolve the ComfyUI install directory (the parent of this `comfy/` package).
-# We deliberately avoid `folder_paths.base_path` here because that is overridden
-# by the `--base-directory` CLI arg to a user-supplied path, whereas the
-# `.comfy_environment` marker is written by launchers/installers next to the
-# ComfyUI install itself.
-_COMFY_INSTALL_DIR = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
-
-
-@functools.cache
-def get_deploy_environment() -> str:
-    env_file = os.path.join(_COMFY_INSTALL_DIR, _ENV_FILENAME)
-    try:
-        with open(env_file, encoding="utf-8") as f:
-            # Cap the read so a malformed or maliciously crafted file (e.g.
-            # a single huge line with no newline) can't blow up memory.
-            first_line = f.readline(128).strip()
-            value = "".join(c for c in first_line if 32 <= ord(c) < 127)
-            if value:
-                return value
-    except FileNotFoundError:
-        pass
-    except Exception as e:
-        logger.error("Failed to read %s: %s", env_file, e)
-
-    return _DEFAULT_DEPLOY_ENV

comfy/hooks.py

@@ -93,7 +93,7 @@ class Hook:
         self.hook_scope = hook_scope
         '''Scope of where this hook should apply in terms of the conds used in sampling run.'''
         self.custom_should_register = default_should_register
-        '''Can be overridden with a compatible function to decide if this hook should be registered without the need to override .should_register'''
+        '''Can be overriden with a compatible function to decide if this hook should be registered without the need to override .should_register'''
 
     @property
     def strength(self):

comfy/k_diffusion/sampling.py

@@ -242,7 +242,6 @@ def sample_euler_ancestral_RF(model, x, sigmas, extra_args=None, callback=None,
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
-    s_noise = s_noise * getattr(model.inner_model.model_patcher.get_model_object('model_sampling'), "noise_scale", 1.0)
     s_in = x.new_ones([x.shape[0]])
     for i in trange(len(sigmas) - 1, disable=disable):
         denoised = model(x, sigmas[i] * s_in, **extra_args)
@@ -374,7 +373,6 @@ def sample_dpm_2_ancestral_RF(model, x, sigmas, extra_args=None, callback=None,
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
-    s_noise = s_noise * getattr(model.inner_model.model_patcher.get_model_object('model_sampling'), "noise_scale", 1.0)
     s_in = x.new_ones([x.shape[0]])
     for i in trange(len(sigmas) - 1, disable=disable):
         denoised = model(x, sigmas[i] * s_in, **extra_args)
@@ -688,7 +686,6 @@ def sample_dpmpp_2s_ancestral_RF(model, x, sigmas, extra_args=None, callback=Non
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
-    s_noise = s_noise * getattr(model.inner_model.model_patcher.get_model_object('model_sampling'), "noise_scale", 1.0)
     s_in = x.new_ones([x.shape[0]])
     sigma_fn = lambda lbda: (lbda.exp() + 1) ** -1
     lambda_fn = lambda sigma: ((1-sigma)/sigma).log()
@@ -750,7 +747,6 @@ def sample_dpmpp_sde(model, x, sigmas, extra_args=None, callback=None, disable=N
     sigma_fn = partial(half_log_snr_to_sigma, model_sampling=model_sampling)
     lambda_fn = partial(sigma_to_half_log_snr, model_sampling=model_sampling)
     sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
 
     for i in trange(len(sigmas) - 1, disable=disable):
         denoised = model(x, sigmas[i] * s_in, **extra_args)
@@ -836,7 +832,6 @@ def sample_dpmpp_2m_sde(model, x, sigmas, extra_args=None, callback=None, disabl
     model_sampling = model.inner_model.model_patcher.get_model_object('model_sampling')
     lambda_fn = partial(sigma_to_half_log_snr, model_sampling=model_sampling)
     sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
 
     old_denoised = None
     h, h_last = None, None
@@ -894,7 +889,6 @@ def sample_dpmpp_3m_sde(model, x, sigmas, extra_args=None, callback=None, disabl
     model_sampling = model.inner_model.model_patcher.get_model_object('model_sampling')
     lambda_fn = partial(sigma_to_half_log_snr, model_sampling=model_sampling)
     sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
 
     denoised_1, denoised_2 = None, None
     h, h_1, h_2 = None, None, None
@@ -1012,39 +1006,23 @@ def sample_ddpm(model, x, sigmas, extra_args=None, callback=None, disable=None,
     return generic_step_sampler(model, x, sigmas, extra_args, callback, disable, noise_sampler, DDPMSampler_step)
 
 @torch.no_grad()
-def sample_lcm(model, x, sigmas, extra_args=None, callback=None, disable=None, noise_sampler=None, s_noise=1.0, s_noise_end=None, noise_clip_std=0.0):
-
-    # s_noise / s_noise_end: per-step noise multiplier, linearly interpolated across steps
-    # noise_clip_std: clamp injected noise to +/- N stddevs (0 disables).
-
+def sample_lcm(model, x, sigmas, extra_args=None, callback=None, disable=None, noise_sampler=None):
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
     s_in = x.new_ones([x.shape[0]])
-    n_steps = max(1, len(sigmas) - 1)
-    model_sampling = model.inner_model.model_patcher.get_model_object('model_sampling')
-
-    s_start = float(s_noise)
-    s_end = s_start if s_noise_end is None else float(s_noise_end)
-    for i in trange(n_steps, disable=disable):
+    for i in trange(len(sigmas) - 1, disable=disable):
         denoised = model(x, sigmas[i] * s_in, **extra_args)
         if callback is not None:
             callback({'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})
 
         x = denoised
         if sigmas[i + 1] > 0:
-            noise = noise_sampler(sigmas[i], sigmas[i + 1])
-            if noise_clip_std > 0:
-                clip_val = noise_clip_std * noise.std()
-                noise = noise.clamp(min=-clip_val, max=clip_val)
-            t = (i / (n_steps - 1)) if n_steps > 1 else 0.0
-            s_noise_i = s_start + (s_end - s_start) * t
-            if s_noise_i != 1.0:
-                noise = noise * s_noise_i
-            x = model_sampling.noise_scaling(sigmas[i + 1], noise, x)
+            x = model.inner_model.inner_model.model_sampling.noise_scaling(sigmas[i + 1], noise_sampler(sigmas[i], sigmas[i + 1]), x)
     return x
 
 
+
 @torch.no_grad()
 def sample_heunpp2(model, x, sigmas, extra_args=None, callback=None, disable=None, s_churn=0., s_tmin=0., s_tmax=float('inf'), s_noise=1.):
     # From MIT licensed: https://github.com/Carzit/sd-webui-samplers-scheduler/
@@ -1271,7 +1249,6 @@ def sample_euler_ancestral_cfg_pp(model, x, sigmas, extra_args=None, callback=No
 
     model_sampling = model.inner_model.model_patcher.get_model_object("model_sampling")
     lambda_fn = partial(sigma_to_half_log_snr, model_sampling=model_sampling)
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
 
     uncond_denoised = None
 
@@ -1319,7 +1296,6 @@ def sample_dpmpp_2s_ancestral_cfg_pp(model, x, sigmas, extra_args=None, callback
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
-    s_noise = s_noise * getattr(model.inner_model.model_patcher.get_model_object('model_sampling'), "noise_scale", 1.0)
 
     temp = [0]
     def post_cfg_function(args):
@@ -1395,7 +1371,6 @@ def res_multistep(model, x, sigmas, extra_args=None, callback=None, disable=None
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
-    s_noise = s_noise * getattr(model.inner_model.model_patcher.get_model_object('model_sampling'), "noise_scale", 1.0)
     s_in = x.new_ones([x.shape[0]])
     sigma_fn = lambda t: t.neg().exp()
     t_fn = lambda sigma: sigma.log().neg()
@@ -1529,7 +1504,6 @@ def sample_er_sde(model, x, sigmas, extra_args=None, callback=None, disable=None
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
-    s_noise = s_noise * getattr(model.inner_model.model_patcher.get_model_object('model_sampling'), "noise_scale", 1.0)
     s_in = x.new_ones([x.shape[0]])
 
     def default_er_sde_noise_scaler(x):
@@ -1600,10 +1574,9 @@ def sample_seeds_2(model, x, sigmas, extra_args=None, callback=None, disable=Non
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
     s_in = x.new_ones([x.shape[0]])
+    inject_noise = eta > 0 and s_noise > 0
 
     model_sampling = model.inner_model.model_patcher.get_model_object('model_sampling')
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
-    inject_noise = eta > 0 and s_noise > 0
     sigma_fn = partial(half_log_snr_to_sigma, model_sampling=model_sampling)
     lambda_fn = partial(sigma_to_half_log_snr, model_sampling=model_sampling)
     sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
@@ -1672,10 +1645,9 @@ def sample_seeds_3(model, x, sigmas, extra_args=None, callback=None, disable=Non
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
     s_in = x.new_ones([x.shape[0]])
+    inject_noise = eta > 0 and s_noise > 0
 
     model_sampling = model.inner_model.model_patcher.get_model_object('model_sampling')
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
-    inject_noise = eta > 0 and s_noise > 0
     sigma_fn = partial(half_log_snr_to_sigma, model_sampling=model_sampling)
     lambda_fn = partial(sigma_to_half_log_snr, model_sampling=model_sampling)
     sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
@@ -1741,7 +1713,6 @@ def sample_sa_solver(model, x, sigmas, extra_args=None, callback=None, disable=F
     s_in = x.new_ones([x.shape[0]])
 
     model_sampling = model.inner_model.model_patcher.get_model_object("model_sampling")
-    s_noise = s_noise * getattr(model_sampling, "noise_scale", 1.0)
     sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
     lambdas = sigma_to_half_log_snr(sigmas, model_sampling=model_sampling)
 
@@ -1839,119 +1810,3 @@ def sample_sa_solver(model, x, sigmas, extra_args=None, callback=None, disable=F
 def sample_sa_solver_pece(model, x, sigmas, extra_args=None, callback=None, disable=False, tau_func=None, s_noise=1.0, noise_sampler=None, predictor_order=3, corrector_order=4, simple_order_2=False):
     """Stochastic Adams Solver with PECE (Predict–Evaluate–Correct–Evaluate) mode (NeurIPS 2023)."""
     return sample_sa_solver(model, x, sigmas, extra_args=extra_args, callback=callback, disable=disable, tau_func=tau_func, s_noise=s_noise, noise_sampler=noise_sampler, predictor_order=predictor_order, corrector_order=corrector_order, use_pece=True, simple_order_2=simple_order_2)
-
-
-@torch.no_grad()
-def sample_ar_video(model, x, sigmas, extra_args=None, callback=None, disable=None,
-                    num_frame_per_block=1):
-    """
-    Autoregressive video sampler: block-by-block denoising with KV cache
-    and flow-match re-noising for Causal Forcing / Self-Forcing models.
-
-    Requires a Causal-WAN compatible model (diffusion_model must expose
-    init_kv_caches / init_crossattn_caches) and 5-D latents [B,C,T,H,W].
-
-    All AR-loop parameters are passed via the SamplerARVideo node, not read
-    from the checkpoint or transformer_options.
-    """
-    extra_args = {} if extra_args is None else extra_args
-    model_options = extra_args.get("model_options", {})
-    transformer_options = model_options.get("transformer_options", {})
-
-    if x.ndim != 5:
-        raise ValueError(
-            f"ar_video sampler requires 5-D video latents [B,C,T,H,W], got {x.ndim}-D tensor with shape {x.shape}. "
-            "This sampler is only compatible with autoregressive video models (e.g. Causal-WAN)."
-        )
-
-    inner_model = model.inner_model.inner_model
-    causal_model = inner_model.diffusion_model
-
-    if not (hasattr(causal_model, "init_kv_caches") and hasattr(causal_model, "init_crossattn_caches")):
-        raise TypeError(
-            "ar_video sampler requires a Causal-WAN compatible model whose diffusion_model "
-            "exposes init_kv_caches() and init_crossattn_caches(). The loaded checkpoint "
-            "does not support this interface — choose a different sampler."
-        )
-
-    seed = extra_args.get("seed", 0)
-
-    bs, c, lat_t, lat_h, lat_w = x.shape
-    frame_seq_len = -(-lat_h // 2) * -(-lat_w // 2) # ceiling division
-    num_blocks = -(-lat_t // num_frame_per_block)   # ceiling division
-    device = x.device
-    model_dtype = inner_model.get_dtype()
-
-    kv_caches = causal_model.init_kv_caches(bs, lat_t * frame_seq_len, device, model_dtype)
-    crossattn_caches = causal_model.init_crossattn_caches(bs, device, model_dtype)
-
-    output = torch.zeros_like(x)
-    s_in = x.new_ones([x.shape[0]])
-    current_start_frame = 0
-
-    # I2V: seed KV cache with the initial image latent before the denoising loop
-    initial_latent = transformer_options.get("ar_config", {}).get("initial_latent", None)
-    if initial_latent is not None:
-        initial_latent = inner_model.process_latent_in(initial_latent).to(device=device, dtype=model_dtype)
-        n_init = initial_latent.shape[2]
-        output[:, :, :n_init] = initial_latent
-
-        ar_state = {"start_frame": 0, "kv_caches": kv_caches, "crossattn_caches": crossattn_caches}
-        transformer_options["ar_state"] = ar_state
-        zero_sigma = sigmas.new_zeros([1])
-        _ = model(initial_latent, zero_sigma * s_in, **extra_args)
-
-        current_start_frame = n_init
-        remaining = lat_t - n_init
-        num_blocks = -(-remaining // num_frame_per_block)
-
-    num_sigma_steps = len(sigmas) - 1
-    total_real_steps = num_blocks * num_sigma_steps
-    step_count = 0
-
-    try:
-        for block_idx in trange(num_blocks, disable=disable):
-            bf = min(num_frame_per_block, lat_t - current_start_frame)
-            fs, fe = current_start_frame, current_start_frame + bf
-            noisy_input = x[:, :, fs:fe]
-
-            ar_state = {
-                "start_frame": current_start_frame,
-                "kv_caches": kv_caches,
-                "crossattn_caches": crossattn_caches,
-            }
-            transformer_options["ar_state"] = ar_state
-
-            for i in range(num_sigma_steps):
-                denoised = model(noisy_input, sigmas[i] * s_in, **extra_args)
-
-                if callback is not None:
-                    scaled_i = step_count * num_sigma_steps // total_real_steps
-                    callback({"x": noisy_input, "i": scaled_i, "sigma": sigmas[i],
-                              "sigma_hat": sigmas[i], "denoised": denoised})
-
-                if sigmas[i + 1] == 0:
-                    noisy_input = denoised
-                else:
-                    sigma_next = sigmas[i + 1]
-                    torch.manual_seed(seed + block_idx * 1000 + i)
-                    fresh_noise = torch.randn_like(denoised)
-                    noisy_input = (1.0 - sigma_next) * denoised + sigma_next * fresh_noise
-
-                    for cache in kv_caches:
-                        cache["end"] -= bf * frame_seq_len
-
-                step_count += 1
-
-            output[:, :, fs:fe] = noisy_input
-
-            for cache in kv_caches:
-                cache["end"] -= bf * frame_seq_len
-            zero_sigma = sigmas.new_zeros([1])
-            _ = model(noisy_input, zero_sigma * s_in, **extra_args)
-
-            current_start_frame += bf
-    finally:
-        transformer_options.pop("ar_state", None)
-
-    return output

comfy/latent_formats.py

@@ -9,7 +9,6 @@ class LatentFormat:
     latent_rgb_factors_reshape = None
     taesd_decoder_name = None
     spacial_downscale_ratio = 8
-    temporal_downscale_ratio = 1
 
     def process_in(self, latent):
         return latent * self.scale_factor
@@ -225,7 +224,6 @@ class Flux2(LatentFormat):
 
         self.latent_rgb_factors_bias = [-0.0329, -0.0718, -0.0851]
         self.latent_rgb_factors_reshape = lambda t: t.reshape(t.shape[0], 32, 2, 2, t.shape[-2], t.shape[-1]).permute(0, 1, 4, 2, 5, 3).reshape(t.shape[0], 32, t.shape[-2] * 2, t.shape[-1] * 2)
-        self.taesd_decoder_name = "taef2_decoder"
 
     def process_in(self, latent):
         return latent
@@ -236,7 +234,6 @@ class Flux2(LatentFormat):
 class Mochi(LatentFormat):
     latent_channels = 12
     latent_dimensions = 3
-    temporal_downscale_ratio = 6
 
     def __init__(self):
         self.scale_factor = 1.0
@@ -280,7 +277,6 @@ class LTXV(LatentFormat):
     latent_channels = 128
     latent_dimensions = 3
     spacial_downscale_ratio = 32
-    temporal_downscale_ratio = 8
 
     def __init__(self):
         self.latent_rgb_factors = [
@@ -424,7 +420,6 @@ class LTXAV(LTXV):
 class HunyuanVideo(LatentFormat):
     latent_channels = 16
     latent_dimensions = 3
-    temporal_downscale_ratio = 4
     scale_factor = 0.476986
     latent_rgb_factors = [
         [-0.0395, -0.0331,  0.0445],
@@ -451,7 +446,6 @@ class HunyuanVideo(LatentFormat):
 class Cosmos1CV8x8x8(LatentFormat):
     latent_channels = 16
     latent_dimensions = 3
-    temporal_downscale_ratio = 8
 
     latent_rgb_factors = [
         [ 0.1817,  0.2284,  0.2423],
@@ -477,7 +471,6 @@ class Cosmos1CV8x8x8(LatentFormat):
 class Wan21(LatentFormat):
     latent_channels = 16
     latent_dimensions = 3
-    temporal_downscale_ratio = 4
 
     latent_rgb_factors = [
             [-0.1299, -0.1692,  0.2932],
@@ -740,7 +733,6 @@ class HunyuanVideo15(LatentFormat):
     latent_channels = 32
     latent_dimensions = 3
     spacial_downscale_ratio = 16
-    temporal_downscale_ratio = 4
     scale_factor = 1.03682
     taesd_decoder_name = "lighttaehy1_5"
 
@@ -791,36 +783,3 @@ class ZImagePixelSpace(ChromaRadiance):
     No VAE encoding/decoding — the model operates directly on RGB pixels.
     """
     pass
-
-
-class HiDreamO1Pixel(ChromaRadiance):
-    """Pixel-space latent format for HiDream-O1.
-    No VAE — model patches/unpatches raw RGB internally with patch_size=32.
-    """
-    pass
-
-class CogVideoX(LatentFormat):
-    """Latent format for CogVideoX-2b (THUDM/CogVideoX-2b).
-
-    scale_factor matches the vae/config.json scaling_factor for the 2b variant.
-    The 5b-class checkpoints (CogVideoX-5b, CogVideoX-1.5-5B, CogVideoX-Fun-V1.5-*)
-    use a different value; see CogVideoX1_5 below.
-    """
-    latent_channels = 16
-    latent_dimensions = 3
-    temporal_downscale_ratio = 4
-
-    def __init__(self):
-        self.scale_factor = 1.15258426
-
-
-class CogVideoX1_5(CogVideoX):
-    """Latent format for 5b-class CogVideoX checkpoints.
-
-    Covers THUDM/CogVideoX-5b, THUDM/CogVideoX-1.5-5B, and the CogVideoX-Fun
-    V1.5-5b family (including VOID inpainting). All of these have
-    scaling_factor=0.7 in their vae/config.json. Auto-selected in
-    supported_models.CogVideoX_T2V based on transformer hidden dim.
-    """
-    def __init__(self):
-        self.scale_factor = 0.7

comfy/ldm/ace/ace_step15.py

@@ -611,7 +611,6 @@ class AceStepDiTModel(nn.Module):
         intermediate_size,
         patch_size,
         audio_acoustic_hidden_dim,
-        condition_dim=None,
         layer_types=None,
         sliding_window=128,
         rms_norm_eps=1e-6,
@@ -641,7 +640,7 @@ class AceStepDiTModel(nn.Module):
 
         self.time_embed = TimestepEmbedding(256, hidden_size, dtype=dtype, device=device, operations=operations)
         self.time_embed_r = TimestepEmbedding(256, hidden_size, dtype=dtype, device=device, operations=operations)
-        self.condition_embedder = Linear(condition_dim, hidden_size, dtype=dtype, device=device)
+        self.condition_embedder = Linear(hidden_size, hidden_size, dtype=dtype, device=device)
 
         if layer_types is None:
             layer_types = ["full_attention"] * num_layers
@@ -1036,9 +1035,6 @@ class AceStepConditionGenerationModel(nn.Module):
         fsq_dim=2048,
         fsq_levels=[8, 8, 8, 5, 5, 5],
         fsq_input_num_quantizers=1,
-        encoder_hidden_size=2048,
-        encoder_intermediate_size=6144,
-        encoder_num_heads=16,
         audio_model=None,
         dtype=None,
         device=None,
@@ -1058,24 +1054,24 @@ class AceStepConditionGenerationModel(nn.Module):
 
         self.decoder = AceStepDiTModel(
             in_channels, hidden_size, num_dit_layers, num_heads, num_kv_heads, head_dim,
-            intermediate_size, patch_size, audio_acoustic_hidden_dim, condition_dim=encoder_hidden_size,
+            intermediate_size, patch_size, audio_acoustic_hidden_dim,
             layer_types=layer_types, sliding_window=sliding_window, rms_norm_eps=rms_norm_eps,
             dtype=dtype, device=device, operations=operations
         )
         self.encoder = AceStepConditionEncoder(
-            text_hidden_dim, timbre_hidden_dim, encoder_hidden_size, num_lyric_layers, num_timbre_layers,
-            encoder_num_heads, num_kv_heads, head_dim, encoder_intermediate_size, rms_norm_eps,
+            text_hidden_dim, timbre_hidden_dim, hidden_size, num_lyric_layers, num_timbre_layers,
+            num_heads, num_kv_heads, head_dim, intermediate_size, rms_norm_eps,
             dtype=dtype, device=device, operations=operations
         )
         self.tokenizer = AceStepAudioTokenizer(
-            audio_acoustic_hidden_dim, encoder_hidden_size, pool_window_size, fsq_dim=fsq_dim, fsq_levels=fsq_levels, fsq_input_num_quantizers=fsq_input_num_quantizers, num_layers=num_tokenizer_layers, head_dim=head_dim, rms_norm_eps=rms_norm_eps,
+            audio_acoustic_hidden_dim, hidden_size, pool_window_size, fsq_dim=fsq_dim, fsq_levels=fsq_levels, fsq_input_num_quantizers=fsq_input_num_quantizers, num_layers=num_tokenizer_layers, head_dim=head_dim, rms_norm_eps=rms_norm_eps,
             dtype=dtype, device=device, operations=operations
         )
         self.detokenizer = AudioTokenDetokenizer(
-            encoder_hidden_size, pool_window_size, audio_acoustic_hidden_dim, num_layers=2, head_dim=head_dim,
+            hidden_size, pool_window_size, audio_acoustic_hidden_dim, num_layers=2, head_dim=head_dim,
             dtype=dtype, device=device, operations=operations
         )
-        self.null_condition_emb = nn.Parameter(torch.empty(1, 1, encoder_hidden_size, dtype=dtype, device=device))
+        self.null_condition_emb = nn.Parameter(torch.empty(1, 1, hidden_size, dtype=dtype, device=device))
 
     def prepare_condition(
         self,

comfy/ldm/cogvideo/model.py

@@ -1,573 +0,0 @@
-# CogVideoX 3D Transformer - ported to ComfyUI native ops
-# Architecture reference: diffusers CogVideoXTransformer3DModel
-# Style reference: comfy/ldm/wan/model.py
-
-import math
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-from comfy.ldm.modules.attention import optimized_attention
-import comfy.patcher_extension
-import comfy.ldm.common_dit
-
-
-def _get_1d_rotary_pos_embed(dim, pos, theta=10000.0):
-    """Returns (cos, sin) each with shape [seq_len, dim].
-
-    Frequencies are computed at dim//2 resolution then repeat_interleaved
-    to full dim, matching CogVideoX's interleaved (real, imag) pair format.
-    """
-    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2, dtype=torch.float32, device=pos.device) / dim))
-    angles = torch.outer(pos.float(), freqs.float())
-    cos = angles.cos().repeat_interleave(2, dim=-1).float()
-    sin = angles.sin().repeat_interleave(2, dim=-1).float()
-    return (cos, sin)
-
-
-def apply_rotary_emb(x, freqs_cos_sin):
-    """Apply CogVideoX rotary embedding to query or key tensor.
-
-    x: [B, heads, seq_len, head_dim]
-    freqs_cos_sin: (cos, sin) each [seq_len, head_dim//2]
-
-    Uses interleaved pair rotation (same as diffusers CogVideoX/Flux).
-    head_dim is reshaped to (-1, 2) pairs, rotated, then flattened back.
-    """
-    cos, sin = freqs_cos_sin
-    cos = cos[None, None, :, :].to(x.device)
-    sin = sin[None, None, :, :].to(x.device)
-
-    # Interleaved pairs: [B, H, S, D] -> [B, H, S, D//2, 2] -> (real, imag)
-    x_real, x_imag = x.reshape(*x.shape[:-1], -1, 2).unbind(-1)
-    x_rotated = torch.stack([-x_imag, x_real], dim=-1).flatten(3)
-
-    return (x.float() * cos + x_rotated.float() * sin).to(x.dtype)
-
-
-def get_timestep_embedding(timesteps, dim, flip_sin_to_cos=True, downscale_freq_shift=0, scale=1, max_period=10000):
-    half = dim // 2
-    freqs = torch.exp(-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32, device=timesteps.device) / half)
-    args = timesteps[:, None].float() * freqs[None] * scale
-    embedding = torch.cat([torch.sin(args), torch.cos(args)], dim=-1)
-    if flip_sin_to_cos:
-        embedding = torch.cat([embedding[:, half:], embedding[:, :half]], dim=-1)
-    if dim % 2:
-        embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
-    return embedding
-
-
-def get_3d_sincos_pos_embed(embed_dim, spatial_size, temporal_size, spatial_interpolation_scale=1.0, temporal_interpolation_scale=1.0, device=None):
-    if isinstance(spatial_size, int):
-        spatial_size = (spatial_size, spatial_size)
-
-    grid_w = torch.arange(spatial_size[0], dtype=torch.float32, device=device) / spatial_interpolation_scale
-    grid_h = torch.arange(spatial_size[1], dtype=torch.float32, device=device) / spatial_interpolation_scale
-    grid_t = torch.arange(temporal_size, dtype=torch.float32, device=device) / temporal_interpolation_scale
-
-    grid_t, grid_h, grid_w = torch.meshgrid(grid_t, grid_h, grid_w, indexing="ij")
-
-    embed_dim_spatial = 2 * (embed_dim // 3)
-    embed_dim_temporal = embed_dim // 3
-
-    pos_embed_spatial = _get_2d_sincos_pos_embed(embed_dim_spatial, grid_h, grid_w, device=device)
-    pos_embed_temporal = _get_1d_sincos_pos_embed(embed_dim_temporal, grid_t[:, 0, 0], device=device)
-
-    T, H, W = grid_t.shape
-    pos_embed_temporal = pos_embed_temporal.unsqueeze(1).unsqueeze(1).expand(-1, H, W, -1)
-    pos_embed = torch.cat([pos_embed_temporal, pos_embed_spatial], dim=-1)
-
-    return pos_embed
-
-
-def _get_2d_sincos_pos_embed(embed_dim, grid_h, grid_w, device=None):
-    T, H, W = grid_h.shape
-    half_dim = embed_dim // 2
-    pos_h = _get_1d_sincos_pos_embed(half_dim, grid_h.reshape(-1), device=device).reshape(T, H, W, half_dim)
-    pos_w = _get_1d_sincos_pos_embed(half_dim, grid_w.reshape(-1), device=device).reshape(T, H, W, half_dim)
-    return torch.cat([pos_h, pos_w], dim=-1)
-
-
-def _get_1d_sincos_pos_embed(embed_dim, pos, device=None):
-    half = embed_dim // 2
-    freqs = torch.exp(-math.log(10000.0) * torch.arange(start=0, end=half, dtype=torch.float32, device=device) / half)
-    args = pos.float().reshape(-1)[:, None] * freqs[None]
-    embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
-    if embed_dim % 2:
-        embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
-    return embedding
-
-
-
-class CogVideoXPatchEmbed(nn.Module):
-    def __init__(self, patch_size=2, patch_size_t=None, in_channels=16, dim=1920,
-                 text_dim=4096, bias=True, sample_width=90, sample_height=60,
-                 sample_frames=49, temporal_compression_ratio=4,
-                 max_text_seq_length=226, spatial_interpolation_scale=1.875,
-                 temporal_interpolation_scale=1.0, use_positional_embeddings=True,
-                 use_learned_positional_embeddings=True,
-                 device=None, dtype=None, operations=None):
-        super().__init__()
-        self.patch_size = patch_size
-        self.patch_size_t = patch_size_t
-        self.dim = dim
-        self.sample_height = sample_height
-        self.sample_width = sample_width
-        self.sample_frames = sample_frames
-        self.temporal_compression_ratio = temporal_compression_ratio
-        self.max_text_seq_length = max_text_seq_length
-        self.spatial_interpolation_scale = spatial_interpolation_scale
-        self.temporal_interpolation_scale = temporal_interpolation_scale
-        self.use_positional_embeddings = use_positional_embeddings
-        self.use_learned_positional_embeddings = use_learned_positional_embeddings
-
-        if patch_size_t is None:
-            self.proj = operations.Conv2d(in_channels, dim, kernel_size=patch_size, stride=patch_size, bias=bias, device=device, dtype=dtype)
-        else:
-            self.proj = operations.Linear(in_channels * patch_size * patch_size * patch_size_t, dim, device=device, dtype=dtype)
-
-        self.text_proj = operations.Linear(text_dim, dim, device=device, dtype=dtype)
-
-        if use_positional_embeddings or use_learned_positional_embeddings:
-            persistent = use_learned_positional_embeddings
-            pos_embedding = self._get_positional_embeddings(sample_height, sample_width, sample_frames)
-            self.register_buffer("pos_embedding", pos_embedding, persistent=persistent)
-
-    def _get_positional_embeddings(self, sample_height, sample_width, sample_frames, device=None):
-        post_patch_height = sample_height // self.patch_size
-        post_patch_width = sample_width // self.patch_size
-        post_time_compression_frames = (sample_frames - 1) // self.temporal_compression_ratio + 1
-        if self.patch_size_t is not None:
-            post_time_compression_frames = post_time_compression_frames // self.patch_size_t
-        num_patches = post_patch_height * post_patch_width * post_time_compression_frames
-
-        pos_embedding = get_3d_sincos_pos_embed(
-            self.dim,
-            (post_patch_width, post_patch_height),
-            post_time_compression_frames,
-            self.spatial_interpolation_scale,
-            self.temporal_interpolation_scale,
-            device=device,
-        )
-        pos_embedding = pos_embedding.reshape(-1, self.dim)
-        joint_pos_embedding = pos_embedding.new_zeros(
-            1, self.max_text_seq_length + num_patches, self.dim, requires_grad=False
-        )
-        joint_pos_embedding.data[:, self.max_text_seq_length:].copy_(pos_embedding)
-        return joint_pos_embedding
-
-    def forward(self, text_embeds, image_embeds):
-        input_dtype = text_embeds.dtype
-        text_embeds = self.text_proj(text_embeds.to(self.text_proj.weight.dtype)).to(input_dtype)
-        batch_size, num_frames, channels, height, width = image_embeds.shape
-
-        proj_dtype = self.proj.weight.dtype
-        if self.patch_size_t is None:
-            image_embeds = image_embeds.reshape(-1, channels, height, width)
-            image_embeds = self.proj(image_embeds.to(proj_dtype)).to(input_dtype)
-            image_embeds = image_embeds.view(batch_size, num_frames, *image_embeds.shape[1:])
-            image_embeds = image_embeds.flatten(3).transpose(2, 3)
-            image_embeds = image_embeds.flatten(1, 2)
-        else:
-            p = self.patch_size
-            p_t = self.patch_size_t
-            image_embeds = image_embeds.permute(0, 1, 3, 4, 2)
-            image_embeds = image_embeds.reshape(
-                batch_size, num_frames // p_t, p_t, height // p, p, width // p, p, channels
-            )
-            image_embeds = image_embeds.permute(0, 1, 3, 5, 7, 2, 4, 6).flatten(4, 7).flatten(1, 3)
-            image_embeds = self.proj(image_embeds.to(proj_dtype)).to(input_dtype)
-
-        embeds = torch.cat([text_embeds, image_embeds], dim=1).contiguous()
-
-        if self.use_positional_embeddings or self.use_learned_positional_embeddings:
-            text_seq_length = text_embeds.shape[1]
-            num_image_patches = image_embeds.shape[1]
-
-            if self.use_learned_positional_embeddings:
-                image_pos = self.pos_embedding[
-                    :, self.max_text_seq_length:self.max_text_seq_length + num_image_patches
-                ].to(device=embeds.device, dtype=embeds.dtype)
-            else:
-                image_pos = get_3d_sincos_pos_embed(
-                    self.dim,
-                    (width // self.patch_size, height // self.patch_size),
-                    num_image_patches // ((height // self.patch_size) * (width // self.patch_size)),
-                    self.spatial_interpolation_scale,
-                    self.temporal_interpolation_scale,
-                    device=embeds.device,
-                ).reshape(1, num_image_patches, self.dim).to(dtype=embeds.dtype)
-
-            # Build joint: zeros for text + sincos for image
-            joint_pos = torch.zeros(1, text_seq_length + num_image_patches, self.dim, device=embeds.device, dtype=embeds.dtype)
-            joint_pos[:, text_seq_length:] = image_pos
-            embeds = embeds + joint_pos
-
-        return embeds
-
-
-class CogVideoXLayerNormZero(nn.Module):
-    def __init__(self, time_dim, dim, elementwise_affine=True, eps=1e-5, bias=True,
-                 device=None, dtype=None, operations=None):
-        super().__init__()
-        self.silu = nn.SiLU()
-        self.linear = operations.Linear(time_dim, 6 * dim, bias=bias, device=device, dtype=dtype)
-        self.norm = operations.LayerNorm(dim, eps=eps, elementwise_affine=elementwise_affine, device=device, dtype=dtype)
-
-    def forward(self, hidden_states, encoder_hidden_states, temb):
-        shift, scale, gate, enc_shift, enc_scale, enc_gate = self.linear(self.silu(temb)).chunk(6, dim=1)
-        hidden_states = self.norm(hidden_states) * (1 + scale)[:, None, :] + shift[:, None, :]
-        encoder_hidden_states = self.norm(encoder_hidden_states) * (1 + enc_scale)[:, None, :] + enc_shift[:, None, :]
-        return hidden_states, encoder_hidden_states, gate[:, None, :], enc_gate[:, None, :]
-
-
-class CogVideoXAdaLayerNorm(nn.Module):
-    def __init__(self, time_dim, dim, elementwise_affine=True, eps=1e-5,
-                 device=None, dtype=None, operations=None):
-        super().__init__()
-        self.silu = nn.SiLU()
-        self.linear = operations.Linear(time_dim, 2 * dim, device=device, dtype=dtype)
-        self.norm = operations.LayerNorm(dim, eps=eps, elementwise_affine=elementwise_affine, device=device, dtype=dtype)
-
-    def forward(self, x, temb):
-        temb = self.linear(self.silu(temb))
-        shift, scale = temb.chunk(2, dim=1)
-        x = self.norm(x) * (1 + scale)[:, None, :] + shift[:, None, :]
-        return x
-
-
-class CogVideoXBlock(nn.Module):
-    def __init__(self, dim, num_heads, head_dim, time_dim,
-                 eps=1e-5, ff_inner_dim=None, ff_bias=True,
-                 device=None, dtype=None, operations=None):
-        super().__init__()
-        self.dim = dim
-        self.num_heads = num_heads
-        self.head_dim = head_dim
-
-        self.norm1 = CogVideoXLayerNormZero(time_dim, dim, eps=eps, device=device, dtype=dtype, operations=operations)
-
-        # Self-attention (joint text + latent)
-        self.q = operations.Linear(dim, dim, bias=True, device=device, dtype=dtype)
-        self.k = operations.Linear(dim, dim, bias=True, device=device, dtype=dtype)
-        self.v = operations.Linear(dim, dim, bias=True, device=device, dtype=dtype)
-        self.norm_q = operations.LayerNorm(head_dim, eps=1e-6, elementwise_affine=True, device=device, dtype=dtype)
-        self.norm_k = operations.LayerNorm(head_dim, eps=1e-6, elementwise_affine=True, device=device, dtype=dtype)
-        self.attn_out = operations.Linear(dim, dim, bias=True, device=device, dtype=dtype)
-
-        self.norm2 = CogVideoXLayerNormZero(time_dim, dim, eps=eps, device=device, dtype=dtype, operations=operations)
-
-        # Feed-forward (GELU approximate)
-        inner_dim = ff_inner_dim or dim * 4
-        self.ff_proj = operations.Linear(dim, inner_dim, bias=ff_bias, device=device, dtype=dtype)
-        self.ff_out = operations.Linear(inner_dim, dim, bias=ff_bias, device=device, dtype=dtype)
-
-    def forward(self, hidden_states, encoder_hidden_states, temb, image_rotary_emb=None, transformer_options=None):
-        if transformer_options is None:
-            transformer_options = {}
-        text_seq_length = encoder_hidden_states.size(1)
-
-        # Norm & modulate
-        norm_hidden, norm_encoder, gate_msa, enc_gate_msa = self.norm1(hidden_states, encoder_hidden_states, temb)
-
-        # Joint self-attention
-        qkv_input = torch.cat([norm_encoder, norm_hidden], dim=1)
-        b, s, _ = qkv_input.shape
-        n, d = self.num_heads, self.head_dim
-
-        q = self.q(qkv_input).view(b, s, n, d)
-        k = self.k(qkv_input).view(b, s, n, d)
-        v = self.v(qkv_input)
-
-        q = self.norm_q(q).view(b, s, n, d)
-        k = self.norm_k(k).view(b, s, n, d)
-
-        # Apply rotary embeddings to image tokens only (diffusers format: [B, heads, seq, head_dim])
-        if image_rotary_emb is not None:
-            q_img = q[:, text_seq_length:].transpose(1, 2)  # [B, heads, img_seq, head_dim]
-            k_img = k[:, text_seq_length:].transpose(1, 2)
-            q_img = apply_rotary_emb(q_img, image_rotary_emb)
-            k_img = apply_rotary_emb(k_img, image_rotary_emb)
-            q = torch.cat([q[:, :text_seq_length], q_img.transpose(1, 2)], dim=1)
-            k = torch.cat([k[:, :text_seq_length], k_img.transpose(1, 2)], dim=1)
-
-        attn_out = optimized_attention(
-            q.reshape(b, s, n * d),
-            k.reshape(b, s, n * d),
-            v,
-            heads=self.num_heads,
-            transformer_options=transformer_options,
-        )
-
-        attn_out = self.attn_out(attn_out)
-
-        attn_encoder, attn_hidden = attn_out.split([text_seq_length, s - text_seq_length], dim=1)
-
-        hidden_states = hidden_states + gate_msa * attn_hidden
-        encoder_hidden_states = encoder_hidden_states + enc_gate_msa * attn_encoder
-
-        # Norm & modulate for FF
-        norm_hidden, norm_encoder, gate_ff, enc_gate_ff = self.norm2(hidden_states, encoder_hidden_states, temb)
-
-        # Feed-forward (GELU on concatenated text + latent)
-        ff_input = torch.cat([norm_encoder, norm_hidden], dim=1)
-        ff_output = self.ff_out(F.gelu(self.ff_proj(ff_input), approximate="tanh"))
-
-        hidden_states = hidden_states + gate_ff * ff_output[:, text_seq_length:]
-        encoder_hidden_states = encoder_hidden_states + enc_gate_ff * ff_output[:, :text_seq_length]
-
-        return hidden_states, encoder_hidden_states
-
-
-class CogVideoXTransformer3DModel(nn.Module):
-    def __init__(self,
-                 num_attention_heads=30,
-                 attention_head_dim=64,
-                 in_channels=16,
-                 out_channels=16,
-                 flip_sin_to_cos=True,
-                 freq_shift=0,
-                 time_embed_dim=512,
-                 ofs_embed_dim=None,
-                 text_embed_dim=4096,
-                 num_layers=30,
-                 dropout=0.0,
-                 attention_bias=True,
-                 sample_width=90,
-                 sample_height=60,
-                 sample_frames=49,
-                 patch_size=2,
-                 patch_size_t=None,
-                 temporal_compression_ratio=4,
-                 max_text_seq_length=226,
-                 spatial_interpolation_scale=1.875,
-                 temporal_interpolation_scale=1.0,
-                 use_rotary_positional_embeddings=False,
-                 use_learned_positional_embeddings=False,
-                 patch_bias=True,
-                 image_model=None,
-                 device=None,
-                 dtype=None,
-                 operations=None,
-                 ):
-        super().__init__()
-        self.dtype = dtype
-        dim = num_attention_heads * attention_head_dim
-        self.dim = dim
-        self.num_attention_heads = num_attention_heads
-        self.attention_head_dim = attention_head_dim
-        self.in_channels = in_channels
-        self.out_channels = out_channels
-        self.patch_size = patch_size
-        self.patch_size_t = patch_size_t
-        self.max_text_seq_length = max_text_seq_length
-        self.use_rotary_positional_embeddings = use_rotary_positional_embeddings
-
-        # 1. Patch embedding
-        self.patch_embed = CogVideoXPatchEmbed(
-            patch_size=patch_size,
-            patch_size_t=patch_size_t,
-            in_channels=in_channels,
-            dim=dim,
-            text_dim=text_embed_dim,
-            bias=patch_bias,
-            sample_width=sample_width,
-            sample_height=sample_height,
-            sample_frames=sample_frames,
-            temporal_compression_ratio=temporal_compression_ratio,
-            max_text_seq_length=max_text_seq_length,
-            spatial_interpolation_scale=spatial_interpolation_scale,
-            temporal_interpolation_scale=temporal_interpolation_scale,
-            use_positional_embeddings=not use_rotary_positional_embeddings,
-            use_learned_positional_embeddings=use_learned_positional_embeddings,
-            device=device, dtype=torch.float32, operations=operations,
-        )
-
-        # 2. Time embedding
-        self.time_proj_dim = dim
-        self.time_proj_flip = flip_sin_to_cos
-        self.time_proj_shift = freq_shift
-        self.time_embedding_linear_1 = operations.Linear(dim, time_embed_dim, device=device, dtype=dtype)
-        self.time_embedding_act = nn.SiLU()
-        self.time_embedding_linear_2 = operations.Linear(time_embed_dim, time_embed_dim, device=device, dtype=dtype)
-
-        # Optional OFS embedding (CogVideoX 1.5 I2V)
-        self.ofs_proj_dim = ofs_embed_dim
-        if ofs_embed_dim:
-            self.ofs_embedding_linear_1 = operations.Linear(ofs_embed_dim, ofs_embed_dim, device=device, dtype=dtype)
-            self.ofs_embedding_act = nn.SiLU()
-            self.ofs_embedding_linear_2 = operations.Linear(ofs_embed_dim, ofs_embed_dim, device=device, dtype=dtype)
-        else:
-            self.ofs_embedding_linear_1 = None
-
-        # 3. Transformer blocks
-        self.blocks = nn.ModuleList([
-            CogVideoXBlock(
-                dim=dim,
-                num_heads=num_attention_heads,
-                head_dim=attention_head_dim,
-                time_dim=time_embed_dim,
-                eps=1e-5,
-                device=device, dtype=dtype, operations=operations,
-            )
-            for _ in range(num_layers)
-        ])
-
-        self.norm_final = operations.LayerNorm(dim, eps=1e-5, elementwise_affine=True, device=device, dtype=dtype)
-
-        # 4. Output
-        self.norm_out = CogVideoXAdaLayerNorm(
-            time_dim=time_embed_dim, dim=dim, eps=1e-5,
-            device=device, dtype=dtype, operations=operations,
-        )
-
-        if patch_size_t is None:
-            output_dim = patch_size * patch_size * out_channels
-        else:
-            output_dim = patch_size * patch_size * patch_size_t * out_channels
-
-        self.proj_out = operations.Linear(dim, output_dim, device=device, dtype=dtype)
-
-        self.spatial_interpolation_scale = spatial_interpolation_scale
-        self.temporal_interpolation_scale = temporal_interpolation_scale
-        self.temporal_compression_ratio = temporal_compression_ratio
-
-    def forward(self, x, timestep, context, ofs=None, transformer_options=None, **kwargs):
-        if transformer_options is None:
-            transformer_options = {}
-        return comfy.patcher_extension.WrapperExecutor.new_class_executor(
-            self._forward,
-            self,
-            comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL, transformer_options)
-        ).execute(x, timestep, context, ofs, transformer_options, **kwargs)
-
-    def _forward(self, x, timestep, context, ofs=None, transformer_options=None, **kwargs):
-        if transformer_options is None:
-            transformer_options = {}
-        # ComfyUI passes [B, C, T, H, W]
-        batch_size, channels, t, h, w = x.shape
-
-        # Pad to patch size (temporal + spatial), same pattern as WAN
-        p_t = self.patch_size_t if self.patch_size_t is not None else 1
-        x = comfy.ldm.common_dit.pad_to_patch_size(x, (p_t, self.patch_size, self.patch_size))
-
-        # CogVideoX expects [B, T, C, H, W]
-        x = x.permute(0, 2, 1, 3, 4)
-        batch_size, num_frames, channels, height, width = x.shape
-
-        # Time embedding
-        t_emb = get_timestep_embedding(timestep, self.time_proj_dim, self.time_proj_flip, self.time_proj_shift)
-        t_emb = t_emb.to(dtype=x.dtype)
-        emb = self.time_embedding_linear_2(self.time_embedding_act(self.time_embedding_linear_1(t_emb)))
-
-        if self.ofs_embedding_linear_1 is not None and ofs is not None:
-            ofs_emb = get_timestep_embedding(ofs, self.ofs_proj_dim, self.time_proj_flip, self.time_proj_shift)
-            ofs_emb = ofs_emb.to(dtype=x.dtype)
-            ofs_emb = self.ofs_embedding_linear_2(self.ofs_embedding_act(self.ofs_embedding_linear_1(ofs_emb)))
-            emb = emb + ofs_emb
-
-        # Patch embedding
-        hidden_states = self.patch_embed(context, x)
-
-        text_seq_length = context.shape[1]
-        encoder_hidden_states = hidden_states[:, :text_seq_length]
-        hidden_states = hidden_states[:, text_seq_length:]
-
-        # Rotary embeddings (if used)
-        image_rotary_emb = None
-        if self.use_rotary_positional_embeddings:
-            post_patch_height = height // self.patch_size
-            post_patch_width = width // self.patch_size
-            if self.patch_size_t is None:
-                post_time = num_frames
-            else:
-                post_time = num_frames // self.patch_size_t
-            image_rotary_emb = self._get_rotary_emb(post_patch_height, post_patch_width, post_time, device=x.device)
-
-        # Transformer blocks
-        for i, block in enumerate(self.blocks):
-            hidden_states, encoder_hidden_states = block(
-                hidden_states=hidden_states,
-                encoder_hidden_states=encoder_hidden_states,
-                temb=emb,
-                image_rotary_emb=image_rotary_emb,
-                transformer_options=transformer_options,
-            )
-
-        hidden_states = self.norm_final(hidden_states)
-
-        # Output projection
-        hidden_states = self.norm_out(hidden_states, temb=emb)
-        hidden_states = self.proj_out(hidden_states)
-
-        # Unpatchify
-        p = self.patch_size
-        p_t = self.patch_size_t
-
-        if p_t is None:
-            output = hidden_states.reshape(batch_size, num_frames, height // p, width // p, -1, p, p)
-            output = output.permute(0, 1, 4, 2, 5, 3, 6).flatten(5, 6).flatten(3, 4)
-        else:
-            output = hidden_states.reshape(
-                batch_size, (num_frames + p_t - 1) // p_t, height // p, width // p, -1, p_t, p, p
-            )
-            output = output.permute(0, 1, 5, 4, 2, 6, 3, 7).flatten(6, 7).flatten(4, 5).flatten(1, 2)
-
-        # Back to ComfyUI format [B, C, T, H, W] and crop padding
-        output = output.permute(0, 2, 1, 3, 4)[:, :, :t, :h, :w]
-        return output
-
-    def _get_rotary_emb(self, h, w, t, device):
-        """Compute CogVideoX 3D rotary positional embeddings.
-
-        For CogVideoX 1.5 (patch_size_t != None): uses "slice" mode — grid positions
-        are integer arange computed at max_size, then sliced to actual size.
-        For CogVideoX 1.0 (patch_size_t == None): uses "linspace" mode with crop coords
-        scaled by spatial_interpolation_scale.
-        """
-        d = self.attention_head_dim
-        dim_t = d // 4
-        dim_h = d // 8 * 3
-        dim_w = d // 8 * 3
-
-        if self.patch_size_t is not None:
-            # CogVideoX 1.5: "slice" mode — positions are simple integer indices
-            # Compute at max(sample_size, actual_size) then slice to actual
-            base_h = self.patch_embed.sample_height // self.patch_size
-            base_w = self.patch_embed.sample_width // self.patch_size
-            max_h = max(base_h, h)
-            max_w = max(base_w, w)
-
-            grid_h = torch.arange(max_h, device=device, dtype=torch.float32)
-            grid_w = torch.arange(max_w, device=device, dtype=torch.float32)
-            grid_t = torch.arange(t, device=device, dtype=torch.float32)
-        else:
-            # CogVideoX 1.0: "linspace" mode with interpolation scale
-            grid_h = torch.linspace(0, h - 1, h, device=device, dtype=torch.float32) * self.spatial_interpolation_scale
-            grid_w = torch.linspace(0, w - 1, w, device=device, dtype=torch.float32) * self.spatial_interpolation_scale
-            grid_t = torch.arange(t, device=device, dtype=torch.float32)
-
-        freqs_t = _get_1d_rotary_pos_embed(dim_t, grid_t)
-        freqs_h = _get_1d_rotary_pos_embed(dim_h, grid_h)
-        freqs_w = _get_1d_rotary_pos_embed(dim_w, grid_w)
-
-        t_cos, t_sin = freqs_t
-        h_cos, h_sin = freqs_h
-        w_cos, w_sin = freqs_w
-
-        # Slice to actual size (for "slice" mode where grids may be larger)
-        t_cos, t_sin = t_cos[:t], t_sin[:t]
-        h_cos, h_sin = h_cos[:h], h_sin[:h]
-        w_cos, w_sin = w_cos[:w], w_sin[:w]
-
-        # Broadcast and concatenate into [T*H*W, head_dim]
-        t_cos = t_cos[:, None, None, :].expand(-1, h, w, -1)
-        t_sin = t_sin[:, None, None, :].expand(-1, h, w, -1)
-        h_cos = h_cos[None, :, None, :].expand(t, -1, w, -1)
-        h_sin = h_sin[None, :, None, :].expand(t, -1, w, -1)
-        w_cos = w_cos[None, None, :, :].expand(t, h, -1, -1)
-        w_sin = w_sin[None, None, :, :].expand(t, h, -1, -1)
-
-        cos = torch.cat([t_cos, h_cos, w_cos], dim=-1).reshape(t * h * w, -1)
-        sin = torch.cat([t_sin, h_sin, w_sin], dim=-1).reshape(t * h * w, -1)
-        return (cos, sin)

comfy/ldm/cogvideo/vae.py

@@ -1,566 +0,0 @@
-# CogVideoX VAE - ported to ComfyUI native ops
-# Architecture reference: diffusers AutoencoderKLCogVideoX
-# Style reference: comfy/ldm/wan/vae.py
-
-import numpy as np
-
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-import comfy.ops
-ops = comfy.ops.disable_weight_init
-
-
-class CausalConv3d(nn.Module):
-    """Causal 3D convolution with temporal padding.
-
-    Uses comfy.ops.Conv3d with autopad='causal_zero' fast path: when input has
-    a single temporal frame and no cache, the 3D conv weight is sliced to act
-    as a 2D conv, avoiding computation on zero-padded temporal dimensions.
-    """
-    def __init__(self, in_channels, out_channels, kernel_size, stride=1, dilation=1, pad_mode="constant"):
-        super().__init__()
-        if isinstance(kernel_size, int):
-            kernel_size = (kernel_size,) * 3
-
-        time_kernel, height_kernel, width_kernel = kernel_size
-        self.time_kernel_size = time_kernel
-        self.pad_mode = pad_mode
-
-        height_pad = (height_kernel - 1) // 2
-        width_pad = (width_kernel - 1) // 2
-        self.time_causal_padding = (width_pad, width_pad, height_pad, height_pad, time_kernel - 1, 0)
-
-        stride = stride if isinstance(stride, tuple) else (stride, 1, 1)
-        dilation = (dilation, 1, 1)
-        self.conv = ops.Conv3d(
-            in_channels, out_channels, kernel_size,
-            stride=stride, dilation=dilation,
-            padding=(0, height_pad, width_pad),
-        )
-
-    def forward(self, x, conv_cache=None):
-        if self.pad_mode == "replicate":
-            x = F.pad(x, self.time_causal_padding, mode="replicate")
-            conv_cache = None
-        else:
-            kernel_t = self.time_kernel_size
-            if kernel_t > 1:
-                if conv_cache is None and x.shape[2] == 1:
-                    # Fast path: single frame, no cache. All temporal padding
-                    # frames are copies of the input (replicate-style), so the
-                    # 3D conv reduces to a 2D conv with summed temporal kernel.
-                    w = comfy.ops.cast_to_input(self.conv.weight, x)
-                    b = comfy.ops.cast_to_input(self.conv.bias, x) if self.conv.bias is not None else None
-                    w2d = w.sum(dim=2, keepdim=True)
-                    out = F.conv3d(x, w2d, b,
-                                   self.conv.stride, self.conv.padding,
-                                   self.conv.dilation, self.conv.groups)
-                    return out, None
-                cached = [conv_cache] if conv_cache is not None else [x[:, :, :1]] * (kernel_t - 1)
-                x = torch.cat(cached + [x], dim=2)
-            conv_cache = x[:, :, -self.time_kernel_size + 1:].clone() if self.time_kernel_size > 1 else None
-
-        out = self.conv(x)
-        return out, conv_cache
-
-
-def _interpolate_zq(zq, target_size):
-    """Interpolate latent z to target (T, H, W), matching CogVideoX's first-frame-special handling."""
-    t = target_size[0]
-    if t > 1 and t % 2 == 1:
-        z_first = F.interpolate(zq[:, :, :1], size=(1, target_size[1], target_size[2]))
-        z_rest = F.interpolate(zq[:, :, 1:], size=(t - 1, target_size[1], target_size[2]))
-        return torch.cat([z_first, z_rest], dim=2)
-    return F.interpolate(zq, size=target_size)
-
-
-class SpatialNorm3D(nn.Module):
-    """Spatially conditioned normalization."""
-    def __init__(self, f_channels, zq_channels, groups=32):
-        super().__init__()
-        self.norm_layer = ops.GroupNorm(num_channels=f_channels, num_groups=groups, eps=1e-6, affine=True)
-        self.conv_y = CausalConv3d(zq_channels, f_channels, kernel_size=1, stride=1)
-        self.conv_b = CausalConv3d(zq_channels, f_channels, kernel_size=1, stride=1)
-
-    def forward(self, f, zq, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-
-        if zq.shape[-3:] != f.shape[-3:]:
-            zq = _interpolate_zq(zq, f.shape[-3:])
-
-        conv_y, new_cache["conv_y"] = self.conv_y(zq, conv_cache=conv_cache.get("conv_y"))
-        conv_b, new_cache["conv_b"] = self.conv_b(zq, conv_cache=conv_cache.get("conv_b"))
-
-        return self.norm_layer(f) * conv_y + conv_b, new_cache
-
-
-class ResnetBlock3D(nn.Module):
-    """3D ResNet block with optional spatial norm."""
-    def __init__(self, in_channels, out_channels=None, temb_channels=512, groups=32,
-                 eps=1e-6, act_fn="silu", spatial_norm_dim=None, pad_mode="first"):
-        super().__init__()
-        out_channels = out_channels or in_channels
-        self.in_channels = in_channels
-        self.out_channels = out_channels
-        self.spatial_norm_dim = spatial_norm_dim
-
-        if act_fn == "silu":
-            self.nonlinearity = nn.SiLU()
-        elif act_fn == "swish":
-            self.nonlinearity = nn.SiLU()
-        else:
-            self.nonlinearity = nn.SiLU()
-
-        if spatial_norm_dim is None:
-            self.norm1 = ops.GroupNorm(num_channels=in_channels, num_groups=groups, eps=eps)
-            self.norm2 = ops.GroupNorm(num_channels=out_channels, num_groups=groups, eps=eps)
-        else:
-            self.norm1 = SpatialNorm3D(in_channels, spatial_norm_dim, groups=groups)
-            self.norm2 = SpatialNorm3D(out_channels, spatial_norm_dim, groups=groups)
-
-        self.conv1 = CausalConv3d(in_channels, out_channels, kernel_size=3, pad_mode=pad_mode)
-
-        if temb_channels > 0:
-            self.temb_proj = ops.Linear(temb_channels, out_channels)
-
-        self.conv2 = CausalConv3d(out_channels, out_channels, kernel_size=3, pad_mode=pad_mode)
-
-        if in_channels != out_channels:
-            self.conv_shortcut = ops.Conv3d(in_channels, out_channels, kernel_size=1, stride=1, padding=0)
-        else:
-            self.conv_shortcut = None
-
-    def forward(self, x, temb=None, zq=None, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-        residual = x
-
-        if zq is not None:
-            x, new_cache["norm1"] = self.norm1(x, zq, conv_cache=conv_cache.get("norm1"))
-        else:
-            x = self.norm1(x)
-
-        x = self.nonlinearity(x)
-        x, new_cache["conv1"] = self.conv1(x, conv_cache=conv_cache.get("conv1"))
-
-        if temb is not None and hasattr(self, "temb_proj"):
-            x = x + self.temb_proj(self.nonlinearity(temb))[:, :, None, None, None]
-
-        if zq is not None:
-            x, new_cache["norm2"] = self.norm2(x, zq, conv_cache=conv_cache.get("norm2"))
-        else:
-            x = self.norm2(x)
-
-        x = self.nonlinearity(x)
-        x, new_cache["conv2"] = self.conv2(x, conv_cache=conv_cache.get("conv2"))
-
-        if self.conv_shortcut is not None:
-            residual = self.conv_shortcut(residual)
-
-        return x + residual, new_cache
-
-
-class Downsample3D(nn.Module):
-    """3D downsampling with optional temporal compression."""
-    def __init__(self, in_channels, out_channels, kernel_size=3, stride=2, padding=0, compress_time=False):
-        super().__init__()
-        self.conv = ops.Conv2d(in_channels, out_channels, kernel_size=kernel_size, stride=stride, padding=padding)
-        self.compress_time = compress_time
-
-    def forward(self, x):
-        if self.compress_time:
-            b, c, t, h, w = x.shape
-            x = x.permute(0, 3, 4, 1, 2).reshape(b * h * w, c, t)
-            if t % 2 == 1:
-                x_first, x_rest = x[..., 0], x[..., 1:]
-                if x_rest.shape[-1] > 0:
-                    x_rest = F.avg_pool1d(x_rest, kernel_size=2, stride=2)
-                x = torch.cat([x_first[..., None], x_rest], dim=-1)
-                x = x.reshape(b, h, w, c, x.shape[-1]).permute(0, 3, 4, 1, 2)
-            else:
-                x = F.avg_pool1d(x, kernel_size=2, stride=2)
-                x = x.reshape(b, h, w, c, x.shape[-1]).permute(0, 3, 4, 1, 2)
-
-        pad = (0, 1, 0, 1)
-        x = F.pad(x, pad, mode="constant", value=0)
-        b, c, t, h, w = x.shape
-        x = x.permute(0, 2, 1, 3, 4).reshape(b * t, c, h, w)
-        x = self.conv(x)
-        x = x.reshape(b, t, x.shape[1], x.shape[2], x.shape[3]).permute(0, 2, 1, 3, 4)
-        return x
-
-
-class Upsample3D(nn.Module):
-    """3D upsampling with optional temporal decompression."""
-    def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, padding=1, compress_time=False):
-        super().__init__()
-        self.conv = ops.Conv2d(in_channels, out_channels, kernel_size=kernel_size, stride=stride, padding=padding)
-        self.compress_time = compress_time
-
-    def forward(self, x):
-        if self.compress_time:
-            if x.shape[2] > 1 and x.shape[2] % 2 == 1:
-                x_first, x_rest = x[:, :, 0], x[:, :, 1:]
-                x_first = F.interpolate(x_first, scale_factor=2.0)
-                x_rest = F.interpolate(x_rest, scale_factor=2.0)
-                x = torch.cat([x_first[:, :, None, :, :], x_rest], dim=2)
-            elif x.shape[2] > 1:
-                x = F.interpolate(x, scale_factor=2.0)
-            else:
-                x = x.squeeze(2)
-                x = F.interpolate(x, scale_factor=2.0)
-                x = x[:, :, None, :, :]
-        else:
-            b, c, t, h, w = x.shape
-            x = x.permute(0, 2, 1, 3, 4).reshape(b * t, c, h, w)
-            x = F.interpolate(x, scale_factor=2.0)
-            x = x.reshape(b, t, c, *x.shape[2:]).permute(0, 2, 1, 3, 4)
-
-        b, c, t, h, w = x.shape
-        x = x.permute(0, 2, 1, 3, 4).reshape(b * t, c, h, w)
-        x = self.conv(x)
-        x = x.reshape(b, t, *x.shape[1:]).permute(0, 2, 1, 3, 4)
-        return x
-
-
-class DownBlock3D(nn.Module):
-    def __init__(self, in_channels, out_channels, temb_channels=0, num_layers=1,
-                 eps=1e-6, act_fn="silu", groups=32, add_downsample=True,
-                 compress_time=False, pad_mode="first"):
-        super().__init__()
-        self.resnets = nn.ModuleList([
-            ResnetBlock3D(
-                in_channels=in_channels if i == 0 else out_channels,
-                out_channels=out_channels,
-                temb_channels=temb_channels,
-                groups=groups, eps=eps, act_fn=act_fn, pad_mode=pad_mode,
-            )
-            for i in range(num_layers)
-        ])
-        self.downsamplers = nn.ModuleList([Downsample3D(out_channels, out_channels, compress_time=compress_time)]) if add_downsample else None
-
-    def forward(self, x, temb=None, zq=None, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-        for i, resnet in enumerate(self.resnets):
-            x, new_cache[f"resnet_{i}"] = resnet(x, temb, zq, conv_cache=conv_cache.get(f"resnet_{i}"))
-        if self.downsamplers is not None:
-            for ds in self.downsamplers:
-                x = ds(x)
-        return x, new_cache
-
-
-class MidBlock3D(nn.Module):
-    def __init__(self, in_channels, temb_channels=0, num_layers=1,
-                 eps=1e-6, act_fn="silu", groups=32, spatial_norm_dim=None, pad_mode="first"):
-        super().__init__()
-        self.resnets = nn.ModuleList([
-            ResnetBlock3D(
-                in_channels=in_channels, out_channels=in_channels,
-                temb_channels=temb_channels, groups=groups, eps=eps,
-                act_fn=act_fn, spatial_norm_dim=spatial_norm_dim, pad_mode=pad_mode,
-            )
-            for _ in range(num_layers)
-        ])
-
-    def forward(self, x, temb=None, zq=None, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-        for i, resnet in enumerate(self.resnets):
-            x, new_cache[f"resnet_{i}"] = resnet(x, temb, zq, conv_cache=conv_cache.get(f"resnet_{i}"))
-        return x, new_cache
-
-
-class UpBlock3D(nn.Module):
-    def __init__(self, in_channels, out_channels, temb_channels=0, num_layers=1,
-                 eps=1e-6, act_fn="silu", groups=32, spatial_norm_dim=16,
-                 add_upsample=True, compress_time=False, pad_mode="first"):
-        super().__init__()
-        self.resnets = nn.ModuleList([
-            ResnetBlock3D(
-                in_channels=in_channels if i == 0 else out_channels,
-                out_channels=out_channels,
-                temb_channels=temb_channels, groups=groups, eps=eps,
-                act_fn=act_fn, spatial_norm_dim=spatial_norm_dim, pad_mode=pad_mode,
-            )
-            for i in range(num_layers)
-        ])
-        self.upsamplers = nn.ModuleList([Upsample3D(out_channels, out_channels, compress_time=compress_time)]) if add_upsample else None
-
-    def forward(self, x, temb=None, zq=None, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-        for i, resnet in enumerate(self.resnets):
-            x, new_cache[f"resnet_{i}"] = resnet(x, temb, zq, conv_cache=conv_cache.get(f"resnet_{i}"))
-        if self.upsamplers is not None:
-            for us in self.upsamplers:
-                x = us(x)
-        return x, new_cache
-
-
-class Encoder3D(nn.Module):
-    def __init__(self, in_channels=3, out_channels=16,
-                 block_out_channels=(128, 256, 256, 512),
-                 layers_per_block=3, act_fn="silu",
-                 eps=1e-6, groups=32, pad_mode="first",
-                 temporal_compression_ratio=4):
-        super().__init__()
-        temporal_compress_level = int(np.log2(temporal_compression_ratio))
-
-        self.conv_in = CausalConv3d(in_channels, block_out_channels[0], kernel_size=3, pad_mode=pad_mode)
-
-        self.down_blocks = nn.ModuleList()
-        output_channel = block_out_channels[0]
-        for i in range(len(block_out_channels)):
-            input_channel = output_channel
-            output_channel = block_out_channels[i]
-            is_final = i == len(block_out_channels) - 1
-            compress_time = i < temporal_compress_level
-
-            self.down_blocks.append(DownBlock3D(
-                in_channels=input_channel, out_channels=output_channel,
-                temb_channels=0, num_layers=layers_per_block,
-                eps=eps, act_fn=act_fn, groups=groups,
-                add_downsample=not is_final, compress_time=compress_time,
-            ))
-
-        self.mid_block = MidBlock3D(
-            in_channels=block_out_channels[-1], temb_channels=0,
-            num_layers=2, eps=eps, act_fn=act_fn, groups=groups, pad_mode=pad_mode,
-        )
-
-        self.norm_out = ops.GroupNorm(groups, block_out_channels[-1], eps=1e-6)
-        self.conv_act = nn.SiLU()
-        self.conv_out = CausalConv3d(block_out_channels[-1], 2 * out_channels, kernel_size=3, pad_mode=pad_mode)
-
-    def forward(self, x, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-
-        x, new_cache["conv_in"] = self.conv_in(x, conv_cache=conv_cache.get("conv_in"))
-
-        for i, block in enumerate(self.down_blocks):
-            key = f"down_block_{i}"
-            x, new_cache[key] = block(x, None, None, conv_cache.get(key))
-
-        x, new_cache["mid_block"] = self.mid_block(x, None, None, conv_cache=conv_cache.get("mid_block"))
-
-        x = self.norm_out(x)
-        x = self.conv_act(x)
-        x, new_cache["conv_out"] = self.conv_out(x, conv_cache=conv_cache.get("conv_out"))
-
-        return x, new_cache
-
-
-class Decoder3D(nn.Module):
-    def __init__(self, in_channels=16, out_channels=3,
-                 block_out_channels=(128, 256, 256, 512),
-                 layers_per_block=3, act_fn="silu",
-                 eps=1e-6, groups=32, pad_mode="first",
-                 temporal_compression_ratio=4):
-        super().__init__()
-        reversed_channels = list(reversed(block_out_channels))
-        temporal_compress_level = int(np.log2(temporal_compression_ratio))
-
-        self.conv_in = CausalConv3d(in_channels, reversed_channels[0], kernel_size=3, pad_mode=pad_mode)
-
-        self.mid_block = MidBlock3D(
-            in_channels=reversed_channels[0], temb_channels=0,
-            num_layers=2, eps=eps, act_fn=act_fn, groups=groups,
-            spatial_norm_dim=in_channels, pad_mode=pad_mode,
-        )
-
-        self.up_blocks = nn.ModuleList()
-        output_channel = reversed_channels[0]
-        for i in range(len(block_out_channels)):
-            prev_channel = output_channel
-            output_channel = reversed_channels[i]
-            is_final = i == len(block_out_channels) - 1
-            compress_time = i < temporal_compress_level
-
-            self.up_blocks.append(UpBlock3D(
-                in_channels=prev_channel, out_channels=output_channel,
-                temb_channels=0, num_layers=layers_per_block + 1,
-                eps=eps, act_fn=act_fn, groups=groups,
-                spatial_norm_dim=in_channels,
-                add_upsample=not is_final, compress_time=compress_time,
-            ))
-
-        self.norm_out = SpatialNorm3D(reversed_channels[-1], in_channels, groups=groups)
-        self.conv_act = nn.SiLU()
-        self.conv_out = CausalConv3d(reversed_channels[-1], out_channels, kernel_size=3, pad_mode=pad_mode)
-
-    def forward(self, sample, conv_cache=None):
-        new_cache = {}
-        conv_cache = conv_cache or {}
-
-        x, new_cache["conv_in"] = self.conv_in(sample, conv_cache=conv_cache.get("conv_in"))
-
-        x, new_cache["mid_block"] = self.mid_block(x, None, sample, conv_cache=conv_cache.get("mid_block"))
-
-        for i, block in enumerate(self.up_blocks):
-            key = f"up_block_{i}"
-            x, new_cache[key] = block(x, None, sample, conv_cache=conv_cache.get(key))
-
-        x, new_cache["norm_out"] = self.norm_out(x, sample, conv_cache=conv_cache.get("norm_out"))
-        x = self.conv_act(x)
-        x, new_cache["conv_out"] = self.conv_out(x, conv_cache=conv_cache.get("conv_out"))
-
-        return x, new_cache
-
-
-
-class AutoencoderKLCogVideoX(nn.Module):
-    """CogVideoX VAE. Spatial tiling/slicing handled by ComfyUI's VAE wrapper.
-
-    Uses rolling temporal decode: conv_in + mid_block + temporal up_blocks run
-    on the full (low-res) tensor, then the expensive spatial-only up_blocks +
-    norm_out + conv_out are processed in small temporal chunks with conv_cache
-    carrying causal state between chunks. This keeps peak VRAM proportional to
-    chunk_size rather than total frame count.
-    """
-
-    def __init__(self,
-                 in_channels=3, out_channels=3,
-                 block_out_channels=(128, 256, 256, 512),
-                 latent_channels=16, layers_per_block=3,
-                 act_fn="silu", eps=1e-6, groups=32,
-                 temporal_compression_ratio=4,
-                 ):
-        super().__init__()
-        self.latent_channels = latent_channels
-        self.temporal_compression_ratio = temporal_compression_ratio
-
-        self.encoder = Encoder3D(
-            in_channels=in_channels, out_channels=latent_channels,
-            block_out_channels=block_out_channels, layers_per_block=layers_per_block,
-            act_fn=act_fn, eps=eps, groups=groups,
-            temporal_compression_ratio=temporal_compression_ratio,
-        )
-        self.decoder = Decoder3D(
-            in_channels=latent_channels, out_channels=out_channels,
-            block_out_channels=block_out_channels, layers_per_block=layers_per_block,
-            act_fn=act_fn, eps=eps, groups=groups,
-            temporal_compression_ratio=temporal_compression_ratio,
-        )
-
-        self.num_latent_frames_batch_size = 2
-        self.num_sample_frames_batch_size = 8
-
-    def encode(self, x):
-        t = x.shape[2]
-        frame_batch = self.num_sample_frames_batch_size
-        remainder = t % frame_batch
-        conv_cache = None
-        enc = []
-
-        # Process remainder frames first so only the first chunk can have an
-        # odd temporal dimension — where Downsample3D's first-frame-special
-        # handling in temporal compression is actually correct.
-        if remainder > 0:
-            chunk, conv_cache = self.encoder(x[:, :, :remainder], conv_cache=conv_cache)
-            enc.append(chunk.to(x.device))
-
-        for start in range(remainder, t, frame_batch):
-            chunk, conv_cache = self.encoder(x[:, :, start:start + frame_batch], conv_cache=conv_cache)
-            enc.append(chunk.to(x.device))
-
-        enc = torch.cat(enc, dim=2)
-        mean, _ = enc.chunk(2, dim=1)
-        return mean
-
-    def decode(self, z):
-        return self._decode_rolling(z)
-
-    def _decode_batched(self, z):
-        """Original batched decode - processes 2 latent frames through full decoder."""
-        t = z.shape[2]
-        frame_batch = self.num_latent_frames_batch_size
-        num_batches = max(t // frame_batch, 1)
-        conv_cache = None
-        dec = []
-        for i in range(num_batches):
-            remaining = t % frame_batch
-            start = frame_batch * i + (0 if i == 0 else remaining)
-            end = frame_batch * (i + 1) + remaining
-            chunk, conv_cache = self.decoder(z[:, :, start:end], conv_cache=conv_cache)
-            dec.append(chunk.cpu())
-        return torch.cat(dec, dim=2).to(z.device)
-
-    def _decode_rolling(self, z):
-        """Rolling decode - processes low-res layers on full tensor, then rolls
-        through expensive high-res layers in temporal chunks."""
-        decoder = self.decoder
-        device = z.device
-
-        # Determine which up_blocks have temporal upsample vs spatial-only.
-        # Temporal up_blocks are cheap (low res), spatial-only are expensive.
-        temporal_compress_level = int(np.log2(self.temporal_compression_ratio))
-        split_at = temporal_compress_level  # first N up_blocks do temporal upsample
-
-        # Phase 1: conv_in + mid_block + temporal up_blocks on full tensor (low/medium res)
-        x, _ = decoder.conv_in(z)
-        x, _ = decoder.mid_block(x, None, z)
-
-        for i in range(split_at):
-            x, _ = decoder.up_blocks[i](x, None, z)
-
-        # Phase 2: remaining spatial-only up_blocks + norm_out + conv_out in temporal chunks
-        remaining_blocks = list(range(split_at, len(decoder.up_blocks)))
-        chunk_size = 4  # pixel frames per chunk through high-res layers
-        t_expanded = x.shape[2]
-
-        if t_expanded <= chunk_size or len(remaining_blocks) == 0:
-            # Small enough to process in one go
-            for i in remaining_blocks:
-                x, _ = decoder.up_blocks[i](x, None, z)
-            x, _ = decoder.norm_out(x, z)
-            x = decoder.conv_act(x)
-            x, _ = decoder.conv_out(x)
-            return x
-
-        # Expand z temporally once to match Phase 2's time dimension.
-        # z stays at latent spatial resolution so this is small (~16 MB vs ~1.3 GB
-        # for the old approach of pre-interpolating to every pixel resolution).
-        z_time_expanded = _interpolate_zq(z, (t_expanded, z.shape[3], z.shape[4]))
-
-        # Process in temporal chunks, interpolating spatially per-chunk to avoid
-        # allocating full [B, C, t_expanded, H, W] tensors at each resolution.
-        dec_out = []
-        conv_caches = {}
-
-        for chunk_start in range(0, t_expanded, chunk_size):
-            chunk_end = min(chunk_start + chunk_size, t_expanded)
-            x_chunk = x[:, :, chunk_start:chunk_end]
-            z_t_chunk = z_time_expanded[:, :, chunk_start:chunk_end]
-            z_spatial_cache = {}
-
-            for i in remaining_blocks:
-                block = decoder.up_blocks[i]
-                cache_key = f"up_block_{i}"
-                hw_key = (x_chunk.shape[3], x_chunk.shape[4])
-                if hw_key not in z_spatial_cache:
-                    if z_t_chunk.shape[3] == hw_key[0] and z_t_chunk.shape[4] == hw_key[1]:
-                        z_spatial_cache[hw_key] = z_t_chunk
-                    else:
-                        z_spatial_cache[hw_key] = F.interpolate(z_t_chunk, size=(z_t_chunk.shape[2], hw_key[0], hw_key[1]))
-                x_chunk, new_cache = block(x_chunk, None, z_spatial_cache[hw_key], conv_cache=conv_caches.get(cache_key))
-                conv_caches[cache_key] = new_cache
-
-            hw_key = (x_chunk.shape[3], x_chunk.shape[4])
-            if hw_key not in z_spatial_cache:
-                z_spatial_cache[hw_key] = F.interpolate(z_t_chunk, size=(z_t_chunk.shape[2], hw_key[0], hw_key[1]))
-            x_chunk, new_cache = decoder.norm_out(x_chunk, z_spatial_cache[hw_key], conv_cache=conv_caches.get("norm_out"))
-            conv_caches["norm_out"] = new_cache
-            x_chunk = decoder.conv_act(x_chunk)
-            x_chunk, new_cache = decoder.conv_out(x_chunk, conv_cache=conv_caches.get("conv_out"))
-            conv_caches["conv_out"] = new_cache
-
-            dec_out.append(x_chunk.cpu())
-            del z_spatial_cache
-
-        del x, z_time_expanded
-        return torch.cat(dec_out, dim=2).to(device)

comfy/ldm/ernie/model.py

@@ -1,301 +0,0 @@
-import math
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-from comfy.ldm.modules.attention import optimized_attention
-import comfy.model_management
-
-def rope(pos: torch.Tensor, dim: int, theta: int) -> torch.Tensor:
-    assert dim % 2 == 0
-    if not comfy.model_management.supports_fp64(pos.device):
-        device = torch.device("cpu")
-    else:
-        device = pos.device
-
-    scale = torch.arange(0, dim, 2, dtype=torch.float64, device=device) / dim
-    omega = 1.0 / (theta**scale)
-    out = torch.einsum("...n,d->...nd", pos.to(device), omega)
-    out = torch.stack([torch.cos(out), torch.sin(out)], dim=0)
-    return out.to(dtype=torch.float32, device=pos.device)
-
-def apply_rotary_emb(x_in: torch.Tensor, freqs_cis: torch.Tensor) -> torch.Tensor:
-    rot_dim = freqs_cis.shape[-1]
-    x, x_pass = x_in[..., :rot_dim], x_in[..., rot_dim:]
-    cos_ = freqs_cis[0]
-    sin_ = freqs_cis[1]
-    x1, x2 = x.chunk(2, dim=-1)
-    x_rotated = torch.cat((-x2, x1), dim=-1)
-    return torch.cat((x * cos_ + x_rotated * sin_, x_pass), dim=-1)
-
-class ErnieImageEmbedND3(nn.Module):
-    def __init__(self, dim: int, theta: int, axes_dim: tuple):
-        super().__init__()
-        self.dim = dim
-        self.theta = theta
-        self.axes_dim = list(axes_dim)
-
-    def forward(self, ids: torch.Tensor) -> torch.Tensor:
-        emb = torch.cat([rope(ids[..., i], self.axes_dim[i], self.theta) for i in range(3)], dim=-1)
-        emb = emb.unsqueeze(3)  # [2, B, S, 1, head_dim//2]
-        return torch.stack([emb, emb], dim=-1).reshape(*emb.shape[:-1], -1)  # [B, S, 1, head_dim]
-
-class ErnieImagePatchEmbedDynamic(nn.Module):
-    def __init__(self, in_channels: int, embed_dim: int, patch_size: int, operations, device=None, dtype=None):
-        super().__init__()
-        self.patch_size = patch_size
-        self.proj = operations.Conv2d(in_channels, embed_dim, kernel_size=patch_size, stride=patch_size, bias=True, device=device, dtype=dtype)
-
-    def forward(self, x: torch.Tensor) -> torch.Tensor:
-        x = self.proj(x)
-        batch_size, dim, height, width = x.shape
-        return x.reshape(batch_size, dim, height * width).transpose(1, 2).contiguous()
-
-class Timesteps(nn.Module):
-    def __init__(self, num_channels: int, flip_sin_to_cos: bool = False):
-        super().__init__()
-        self.num_channels = num_channels
-        self.flip_sin_to_cos = flip_sin_to_cos
-
-    def forward(self, timesteps: torch.Tensor) -> torch.Tensor:
-        half_dim = self.num_channels // 2
-        exponent = -math.log(10000) * torch.arange(half_dim, dtype=torch.float32, device=timesteps.device) / half_dim
-        emb = torch.exp(exponent)
-        emb = timesteps[:, None].float() * emb[None, :]
-        if self.flip_sin_to_cos:
-            emb = torch.cat([torch.cos(emb), torch.sin(emb)], dim=-1)
-        else:
-            emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=-1)
-        return emb
-
-class TimestepEmbedding(nn.Module):
-    def __init__(self, in_channels: int, time_embed_dim: int, operations, device=None, dtype=None):
-        super().__init__()
-        Linear = operations.Linear
-        self.linear_1 = Linear(in_channels, time_embed_dim, bias=True, device=device, dtype=dtype)
-        self.act = nn.SiLU()
-        self.linear_2 = Linear(time_embed_dim, time_embed_dim, bias=True, device=device, dtype=dtype)
-
-    def forward(self, sample: torch.Tensor) -> torch.Tensor:
-        sample = self.linear_1(sample)
-        sample = self.act(sample)
-        sample = self.linear_2(sample)
-        return sample
-
-class ErnieImageAttention(nn.Module):
-    def __init__(self, query_dim: int, heads: int, dim_head: int, eps: float = 1e-6, operations=None, device=None, dtype=None):
-        super().__init__()
-        self.heads = heads
-        self.head_dim = dim_head
-        self.inner_dim = heads * dim_head
-
-        Linear = operations.Linear
-        RMSNorm = operations.RMSNorm
-
-        self.to_q = Linear(query_dim, self.inner_dim, bias=False, device=device, dtype=dtype)
-        self.to_k = Linear(query_dim, self.inner_dim, bias=False, device=device, dtype=dtype)
-        self.to_v = Linear(query_dim, self.inner_dim, bias=False, device=device, dtype=dtype)
-
-        self.norm_q = RMSNorm(dim_head, eps=eps, elementwise_affine=True, device=device, dtype=dtype)
-        self.norm_k = RMSNorm(dim_head, eps=eps, elementwise_affine=True, device=device, dtype=dtype)
-
-        self.to_out = nn.ModuleList([Linear(self.inner_dim, query_dim, bias=False, device=device, dtype=dtype)])
-
-    def forward(self, x: torch.Tensor, attention_mask: torch.Tensor = None, image_rotary_emb: torch.Tensor = None) -> torch.Tensor:
-        B, S, _ = x.shape
-
-        q_flat = self.to_q(x)
-        k_flat = self.to_k(x)
-        v_flat = self.to_v(x)
-
-        query = q_flat.view(B, S, self.heads, self.head_dim)
-        key = k_flat.view(B, S, self.heads, self.head_dim)
-
-        query = self.norm_q(query)
-        key = self.norm_k(key)
-
-        if image_rotary_emb is not None:
-            query = apply_rotary_emb(query, image_rotary_emb)
-            key = apply_rotary_emb(key, image_rotary_emb)
-
-        q_flat = query.reshape(B, S, -1)
-        k_flat = key.reshape(B, S, -1)
-
-        hidden_states = optimized_attention(q_flat, k_flat, v_flat, self.heads, mask=attention_mask)
-
-        return self.to_out[0](hidden_states)
-
-class ErnieImageFeedForward(nn.Module):
-    def __init__(self, hidden_size: int, ffn_hidden_size: int, operations, device=None, dtype=None):
-        super().__init__()
-        Linear = operations.Linear
-        self.gate_proj = Linear(hidden_size, ffn_hidden_size, bias=False, device=device, dtype=dtype)
-        self.up_proj = Linear(hidden_size, ffn_hidden_size, bias=False, device=device, dtype=dtype)
-        self.linear_fc2 = Linear(ffn_hidden_size, hidden_size, bias=False, device=device, dtype=dtype)
-
-    def forward(self, x: torch.Tensor) -> torch.Tensor:
-        return self.linear_fc2(self.up_proj(x) * F.gelu(self.gate_proj(x)))
-
-class ErnieImageSharedAdaLNBlock(nn.Module):
-    def __init__(self, hidden_size: int, num_heads: int, ffn_hidden_size: int, eps: float = 1e-6, operations=None, device=None, dtype=None):
-        super().__init__()
-        RMSNorm = operations.RMSNorm
-
-        self.adaLN_sa_ln = RMSNorm(hidden_size, eps=eps, device=device, dtype=dtype)
-        self.self_attention = ErnieImageAttention(
-            query_dim=hidden_size,
-            dim_head=hidden_size // num_heads,
-            heads=num_heads,
-            eps=eps,
-            operations=operations,
-            device=device,
-            dtype=dtype
-        )
-        self.adaLN_mlp_ln = RMSNorm(hidden_size, eps=eps, device=device, dtype=dtype)
-        self.mlp = ErnieImageFeedForward(hidden_size, ffn_hidden_size, operations=operations, device=device, dtype=dtype)
-
-    def forward(self, x, rotary_pos_emb, temb, attention_mask=None):
-        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = temb
-
-        residual = x
-        x_norm = self.adaLN_sa_ln(x)
-        x_norm = x_norm * (1 + scale_msa) + shift_msa
-
-        attn_out = self.self_attention(x_norm, attention_mask=attention_mask, image_rotary_emb=rotary_pos_emb)
-        x = residual + gate_msa * attn_out
-
-        residual = x
-        x_norm = self.adaLN_mlp_ln(x)
-        x_norm = x_norm * (1 + scale_mlp) + shift_mlp
-
-        return residual + gate_mlp * self.mlp(x_norm)
-
-class ErnieImageAdaLNContinuous(nn.Module):
-    def __init__(self, hidden_size: int, eps: float = 1e-6, operations=None, device=None, dtype=None):
-        super().__init__()
-        LayerNorm = operations.LayerNorm
-        Linear = operations.Linear
-        self.norm = LayerNorm(hidden_size, elementwise_affine=False, eps=eps, device=device, dtype=dtype)
-        self.linear = Linear(hidden_size, hidden_size * 2, device=device, dtype=dtype)
-
-    def forward(self, x: torch.Tensor, conditioning: torch.Tensor) -> torch.Tensor:
-        scale, shift = self.linear(conditioning).chunk(2, dim=-1)
-        x = self.norm(x)
-        x = torch.addcmul(shift.unsqueeze(1), x, 1 + scale.unsqueeze(1))
-        return x
-
-class ErnieImageModel(nn.Module):
-    def __init__(
-        self,
-        hidden_size: int = 4096,
-        num_attention_heads: int = 32,
-        num_layers: int = 36,
-        ffn_hidden_size: int = 12288,
-        in_channels: int = 128,
-        out_channels: int = 128,
-        patch_size: int = 1,
-        text_in_dim: int = 3072,
-        rope_theta: int = 256,
-        rope_axes_dim: tuple = (32, 48, 48),
-        eps: float = 1e-6,
-        qk_layernorm: bool = True,
-        device=None,
-        dtype=None,
-        operations=None,
-        **kwargs
-    ):
-        super().__init__()
-        self.dtype = dtype
-        self.hidden_size = hidden_size
-        self.num_heads = num_attention_heads
-        self.head_dim = hidden_size // num_attention_heads
-        self.patch_size = patch_size
-        self.out_channels = out_channels
-
-        Linear = operations.Linear
-
-        self.x_embedder = ErnieImagePatchEmbedDynamic(in_channels, hidden_size, patch_size, operations, device, dtype)
-        self.text_proj = Linear(text_in_dim, hidden_size, bias=False, device=device, dtype=dtype) if text_in_dim != hidden_size else None
-
-        self.time_proj = Timesteps(hidden_size, flip_sin_to_cos=False)
-        self.time_embedding = TimestepEmbedding(hidden_size, hidden_size, operations, device, dtype)
-
-        self.pos_embed = ErnieImageEmbedND3(dim=self.head_dim, theta=rope_theta, axes_dim=rope_axes_dim)
-
-        self.adaLN_modulation = nn.Sequential(
-            nn.SiLU(),
-            Linear(hidden_size, 6 * hidden_size, device=device, dtype=dtype)
-        )
-
-        self.layers = nn.ModuleList([
-            ErnieImageSharedAdaLNBlock(hidden_size, num_attention_heads, ffn_hidden_size, eps, operations, device, dtype)
-            for _ in range(num_layers)
-        ])
-
-        self.final_norm = ErnieImageAdaLNContinuous(hidden_size, eps, operations, device, dtype)
-        self.final_linear = Linear(hidden_size, patch_size * patch_size * out_channels, device=device, dtype=dtype)
-
-    def forward(self, x, timesteps, context, **kwargs):
-        device, dtype = x.device, x.dtype
-        B, C, H, W = x.shape
-        p, Hp, Wp = self.patch_size, H // self.patch_size, W // self.patch_size
-        N_img = Hp * Wp
-
-        img_bsh = self.x_embedder(x)
-
-        text_bth = context
-        if self.text_proj is not None and text_bth.numel() > 0:
-            text_bth = self.text_proj(text_bth)
-        Tmax = text_bth.shape[1]
-
-        hidden_states = torch.cat([img_bsh, text_bth], dim=1)
-
-        text_ids = torch.zeros((B, Tmax, 3), device=device, dtype=torch.float32)
-        text_ids[:, :, 0] = torch.linspace(0, Tmax - 1, steps=Tmax, device=x.device, dtype=torch.float32)
-        index = float(Tmax)
-
-        transformer_options = kwargs.get("transformer_options", {})
-        rope_options = transformer_options.get("rope_options", None)
-
-        h_len, w_len = float(Hp), float(Wp)
-        h_offset, w_offset = 0.0, 0.0
-
-        if rope_options is not None:
-            h_len = (h_len - 1.0) * rope_options.get("scale_y", 1.0) + 1.0
-            w_len = (w_len - 1.0) * rope_options.get("scale_x", 1.0) + 1.0
-            index += rope_options.get("shift_t", 0.0)
-            h_offset += rope_options.get("shift_y", 0.0)
-            w_offset += rope_options.get("shift_x", 0.0)
-
-        image_ids = torch.zeros((Hp, Wp, 3), device=device, dtype=torch.float32)
-        image_ids[:, :, 0] = image_ids[:, :, 1] + index
-        image_ids[:, :, 1] = image_ids[:, :, 1] + torch.linspace(h_offset, h_len - 1 + h_offset, steps=Hp, device=device, dtype=torch.float32).unsqueeze(1)
-        image_ids[:, :, 2] = image_ids[:, :, 2] + torch.linspace(w_offset, w_len - 1 + w_offset, steps=Wp, device=device, dtype=torch.float32).unsqueeze(0)
-
-        image_ids = image_ids.view(1, N_img, 3).expand(B, -1, -1)
-
-        rotary_pos_emb = self.pos_embed(torch.cat([image_ids, text_ids], dim=1)).to(x.dtype)
-        del image_ids, text_ids
-
-        sample = self.time_proj(timesteps).to(dtype)
-        c = self.time_embedding(sample)
-
-        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = [
-            t.unsqueeze(1).contiguous() for t in self.adaLN_modulation(c).chunk(6, dim=-1)
-        ]
-
-        temb = [shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp]
-        for layer in self.layers:
-            hidden_states = layer(hidden_states, rotary_pos_emb, temb)
-
-        hidden_states = self.final_norm(hidden_states, c).type_as(hidden_states)
-
-        patches = self.final_linear(hidden_states)[:, :N_img, :]
-        output = (
-            patches.view(B, Hp, Wp, p, p, self.out_channels)
-            .permute(0, 5, 1, 3, 2, 4)
-            .contiguous()
-            .view(B, self.out_channels, H, W)
-        )
-
-        return output

comfy/ldm/flux/math.py

@@ -16,7 +16,7 @@ def attention(q: Tensor, k: Tensor, v: Tensor, pe: Tensor, mask=None, transforme
 
 def rope(pos: Tensor, dim: int, theta: int) -> Tensor:
     assert dim % 2 == 0
-    if not comfy.model_management.supports_fp64(pos.device):
+    if comfy.model_management.is_device_mps(pos.device) or comfy.model_management.is_intel_xpu() or comfy.model_management.is_directml_enabled():
         device = torch.device("cpu")
     else:
         device = pos.device