Use comfy kitchen apply rope in omnigen2 model.

README.md

@@ -382,7 +382,11 @@ For AMD 7600 and maybe other RDNA3 cards: ```HSA_OVERRIDE_GFX_VERSION=11.0.0 pyt
 
 ### AMD ROCm Tips
 
-You can try setting this env variable `PYTORCH_TUNABLEOP_ENABLED=1` which might speed things up at the cost of a very slow initial run.
+You can enable experimental memory efficient attention on recent pytorch in ComfyUI on some AMD GPUs using this command, it should already be enabled by default on RDNA3. If this improves speed for you on latest pytorch on your GPU please report it so that I can enable it by default.
+
+```TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1 python main.py --use-pytorch-cross-attention```
+
+You can also try setting this env variable `PYTORCH_TUNABLEOP_ENABLED=1` which might speed things up at the cost of a very slow initial run.
 
 # Notes
 

comfy_api/latest/_input/video_types.py

@@ -27,13 +27,10 @@ class VideoInput(ABC):
         path: Union[str, IO[bytes]],
         format: VideoContainer = VideoContainer.AUTO,
         codec: VideoCodec = VideoCodec.AUTO,
-        metadata: Optional[dict] = None,
-        bit_depth: int | None = None,
+        metadata: Optional[dict] = None
     ):
         """
         Abstract method to save the video input to a file.
-
-        bit_depth selects the encoded bit depth; None keeps the video's native depth.
         """
         pass
 
@@ -86,14 +83,6 @@ class VideoInput(ABC):
         components = self.get_components()
         return components.images.shape[2], components.images.shape[1]
 
-    def get_bit_depth(self) -> int:
-        """
-        Returns the bit depth of the video (e.g. 8 or 10).
-
-        Default implementation returns 8; subclasses report their real depth.
-        """
-        return 8
-
     def get_duration(self) -> float:
         """
         Returns the duration of the video in seconds.

comfy_api/latest/_input_impl/video_types.py

@@ -52,12 +52,6 @@ def get_open_write_kwargs(
     return open_kwargs
 
 
-def video_stream_bit_depth(stream) -> int:
-    if stream is None or stream.format is None or not stream.format.components:
-        return 8
-    return max(component.bits for component in stream.format.components)
-
-
 class VideoFromFile(VideoInput):
     """
     Class representing video input from a file.
@@ -103,13 +97,6 @@ class VideoFromFile(VideoInput):
                     return stream.width, stream.height
         raise ValueError(f"No video stream found in file '{self.__file}'")
 
-    def get_bit_depth(self) -> int:
-        if isinstance(self.__file, io.BytesIO):
-            self.__file.seek(0)  # Reset the BytesIO object to the beginning
-        with av.open(self.__file, mode="r") as container:
-            video_stream = container.streams.video[0] if len(container.streams.video) > 0 else None
-            return video_stream_bit_depth(video_stream)
-
     def get_duration(self) -> float:
         """
         Returns the duration of the video in seconds.
@@ -270,7 +257,6 @@ class VideoFromFile(VideoInput):
 
         image_format = 'gbrpf32le'
         process_image_format = lambda a: a
-        align_graph = None
         audio = None
 
         streams = [video_stream]
@@ -324,24 +310,7 @@ class VideoFromFile(VideoInput):
 
                             checked_alpha = True
 
-                        # Fix non-deterministic video decode when the video width is not a multiple of 32
-                        # For non-yuvj pixel formats (all H.264/H.265 video)
-                        if image_format in ('gbrpf32le', 'gbrapf32le') and frame.width % 32 != 0:
-                            if align_graph is None:
-                                pad_w = ((frame.width + 31) // 32) * 32
-                                g = av.filter.Graph()
-                                g_src = g.add_buffer(width=frame.width, height=frame.height,
-                                                     format=frame.format.name, time_base=video_stream.time_base)
-                                g_pad = g.add('pad', f'{pad_w}:{frame.height}:0:0')
-                                g_sink = g.add('buffersink')
-                                g_src.link_to(g_pad)
-                                g_pad.link_to(g_sink)
-                                g.configure()
-                                align_graph = (g, g_src, g_sink)
-                            align_graph[1].push(frame)
-                            img = np.ascontiguousarray(align_graph[2].pull().to_ndarray(format=image_format)[:, :frame.width])
-                        else:
-                            img = frame.to_ndarray(format=image_format)
+                        img = frame.to_ndarray(format=image_format)  # shape: (H, W, 4)
                         if frame.rotation != 0:
                             k = int(round(frame.rotation // 90))
                             img = np.rot90(img, k=k, axes=(0, 1)).copy()
@@ -408,32 +377,25 @@ class VideoFromFile(VideoInput):
         format: VideoContainer = VideoContainer.AUTO,
         codec: VideoCodec = VideoCodec.AUTO,
         metadata: Optional[dict] = None,
-        bit_depth: int | None = None,
     ):
         if isinstance(self.__file, io.BytesIO):
             self.__file.seek(0)  # Reset the BytesIO object to the beginning
         with av.open(self.__file, mode='r') as container:
             container_format = container.format.name
-            video_stream = container.streams.video[0] if len(container.streams.video) > 0 else None
-            video_encoding = video_stream.codec.name if video_stream is not None else None
-            source_bit_depth = video_stream_bit_depth(video_stream)
+            video_encoding = container.streams.video[0].codec.name if len(container.streams.video) > 0 else None
             reuse_streams = True
             if format != VideoContainer.AUTO and format not in container_format.split(","):
                 reuse_streams = False
             if codec != VideoCodec.AUTO and codec != video_encoding and video_encoding is not None:
                 reuse_streams = False
-            if bit_depth is not None and video_encoding is not None and bit_depth != source_bit_depth:
-                reuse_streams = False
             if self.__start_time or self.__duration:
                 reuse_streams = False
 
             if not reuse_streams:
-                if bit_depth is None:
-                    bit_depth = source_bit_depth
                 components = self.get_components_internal(container)
                 video = VideoFromComponents(components)
                 return video.save_to(
-                    path, format=format, codec=codec, metadata=metadata, bit_depth=bit_depth,
+                    path, format=format, codec=codec, metadata=metadata
                 )
 
             streams = container.streams
@@ -489,10 +451,8 @@ class VideoFromComponents(VideoInput):
     Class representing video input from tensors.
     """
 
-    def __init__(self, components: VideoComponents, bit_depth: int = 8):
+    def __init__(self, components: VideoComponents):
         self.__components = components
-        # Tensor components have no inherent bit depth; this is the depth used when encoding.
-        self.__bit_depth = bit_depth
 
     def get_components(self) -> VideoComponents:
         return VideoComponents(
@@ -501,26 +461,18 @@ class VideoFromComponents(VideoInput):
             frame_rate=self.__components.frame_rate,
         )
 
-    def get_bit_depth(self) -> int:
-        return self.__bit_depth
-
     def save_to(
         self,
         path: str,
         format: VideoContainer = VideoContainer.AUTO,
         codec: VideoCodec = VideoCodec.AUTO,
         metadata: Optional[dict] = None,
-        bit_depth: int | None = None,
     ):
         """Save the video to a file path or BytesIO buffer."""
         if format != VideoContainer.AUTO and format != VideoContainer.MP4:
             raise ValueError("Only MP4 format is supported for now")
         if codec != VideoCodec.AUTO and codec != VideoCodec.H264:
             raise ValueError("Only H264 codec is supported for now")
-        # None means "use the depth this video was created with" (CreateVideo's choice).
-        if bit_depth is None:
-            bit_depth = self.__bit_depth
-        is_10bit = bit_depth >= 10
         extra_kwargs = {}
         if isinstance(format, VideoContainer) and format != VideoContainer.AUTO:
             extra_kwargs["format"] = format.value
@@ -536,11 +488,10 @@ class VideoFromComponents(VideoInput):
 
             frame_rate = Fraction(round(self.__components.frame_rate * 1000), 1000)
             # Create a video stream
-            pix_fmt = "yuv420p10le" if is_10bit else "yuv420p"
             video_stream = output.add_stream('h264', rate=frame_rate)
             video_stream.width = self.__components.images.shape[2]
             video_stream.height = self.__components.images.shape[1]
-            video_stream.pix_fmt = pix_fmt
+            video_stream.pix_fmt = 'yuv420p'
 
             # Create an audio stream
             audio_sample_rate = 1
@@ -554,14 +505,9 @@ class VideoFromComponents(VideoInput):
 
             # Encode video
             for i, frame in enumerate(self.__components.images):
-                if is_10bit:
-                    # 16-bit RGB keeps float precision through the conversion to 10-bit YUV.
-                    img = (frame.float() * 65535).clamp(0, 65535).cpu().numpy().astype(np.uint16)  # shape: (H, W, 3)
-                    frame = av.VideoFrame.from_ndarray(img, format="rgb48le")
-                else:
-                    img = (frame * 255).clamp(0, 255).byte().cpu().numpy() # shape: (H, W, 3)
-                    frame = av.VideoFrame.from_ndarray(img, format='rgb24')
-                frame = frame.reformat(format=pix_fmt)
+                img = (frame * 255).clamp(0, 255).byte().cpu().numpy() # shape: (H, W, 3)
+                frame = av.VideoFrame.from_ndarray(img, format='rgb24')
+                frame = frame.reformat(format='yuv420p')  # Convert to YUV420P as required by h264
                 packet = video_stream.encode(frame)
                 output.mux(packet)
 

comfy_api/latest/_io.py

@@ -891,14 +891,6 @@ class Tracks(ComfyTypeIO):
         track_visibility: torch.Tensor
     Type = TrackDict
 
-@comfytype(io_type="COMFY_DICT")
-class ComfyDict(ComfyTypeIO):
-    Type = dict
-
-@comfytype(io_type="COMFY_LIST")
-class ComfyList(ComfyTypeIO):
-    Type = list
-
 @comfytype(io_type="COMFY_MULTITYPED_V3")
 class MultiType:
     Type = Any
@@ -1334,32 +1326,6 @@ class Curve(ComfyTypeIO):
             return d
 
 
-@comfytype(io_type="COLORS")
-class Colors(ComfyTypeIO):
-    Type = list[Color.Type]
-
-    class Input(WidgetInput):
-        def __init__(self, id: str, display_name: str=None, optional=False, tooltip: str=None,
-                     socketless: bool=True, default: list[str]=None, advanced: bool=None):
-            super().__init__(id, display_name, optional, tooltip, None, default, socketless, None, None, None, None, advanced)
-            if default is None:
-                self.default = []
-
-
-@comfytype(io_type="BOUNDING_BOXES")
-class BoundingBoxes(ComfyTypeIO):
-    class BoundingBoxWithMetadata(BoundingBox.BoundingBoxDict):
-        metadata: dict
-    Type = list[BoundingBoxWithMetadata]
-
-    class Input(WidgetInput):
-        def __init__(self, id: str, display_name: str=None, optional=False, tooltip: str=None,
-                     socketless: bool=True, default: list[dict]=None, advanced: bool=None):
-            super().__init__(id, display_name, optional, tooltip, None, default, socketless, None, None, None, None, advanced)
-            if default is None:
-                self.default = []
-
-
 @comfytype(io_type="HISTOGRAM")
 class Histogram(ComfyTypeIO):
     """A histogram represented as a list of bin counts."""
@@ -2410,8 +2376,6 @@ __all__ = [
     "AnyType",
     "MultiType",
     "Tracks",
-    "ComfyDict",
-    "ComfyList",
     "Color",
     # Dynamic Types
     "MatchType",
@@ -2430,8 +2394,6 @@ __all__ = [
     "PriceBadgeDepends",
     "PriceBadge",
     "BoundingBox",
-    "BoundingBoxes",
-    "Colors",
     "Curve",
     "Histogram",
     "Range",

comfy_api_nodes/apis/tripo.py

@@ -208,10 +208,6 @@ class TripoMultiviewToModelRequest(BaseModel):
     quad: bool | None = Field(False, description="Whether to apply quad to the generated model")
 
 
-class TripoTexturePrompt(BaseModel):
-    text: str | None = Field(None, description="Text guidance for texture generation")
-
-
 class TripoTextureModelRequest(BaseModel):
     type: TripoTaskType = Field(TripoTaskType.TEXTURE_MODEL, description="Type of task")
     original_model_task_id: str = Field(..., description="The task ID of the original model")
@@ -223,11 +219,6 @@ class TripoTextureModelRequest(BaseModel):
     texture_alignment: TripoTextureAlignment | None = Field(
         TripoTextureAlignment.ORIGINAL_IMAGE, description="The texture alignment method"
     )
-    texture_prompt: TripoTexturePrompt | None = Field(
-        None,
-        description="Optional guidance for texturing. Required in practice for imported models, "
-        "which carry no source image to infer texture from.",
-    )
 
 
 class TripoRefineModelRequest(BaseModel):
@@ -316,17 +307,6 @@ class TripoP1MultiviewToModelRequest(TripoP1CommonRequest):
     orientation: str | None = None
 
 
-class TripoImportModelRequest(BaseModel):
-    """Request for the comfy-api composite import endpoint (/proxy/tripo/v2/openapi/import).
-
-    The model file is uploaded to ComfyUI API storage first; the backend downloads it from
-    `url`, re-uploads it to Tripo's storage and creates the import_model task server-side.
-    """
-
-    url: str = Field(..., description="ComfyUI API storage download URL of the model file")
-    format: str = Field(..., description='File format: "glb", "fbx", "obj" or "stl"')
-
-
 class TripoTaskOutput(BaseModel):
     model: str | None = Field(None, description="URL to the model")
     base_model: str | None = Field(None, description="URL to the base model")

comfy_api_nodes/nodes_tripo.py

@@ -1,6 +1,6 @@
 from typing_extensions import override
 
-from comfy_api.latest import IO, ComfyExtension, Input, Types
+from comfy_api.latest import IO, ComfyExtension, Input
 from comfy_api_nodes.apis.tripo import (
     TripoAnimateRetargetRequest,
     TripoAnimateRigRequest,
@@ -8,7 +8,6 @@ from comfy_api_nodes.apis.tripo import (
     TripoFileEmptyReference,
     TripoFileReference,
     TripoImageToModelRequest,
-    TripoImportModelRequest,
     TripoModelVersion,
     TripoMultiviewToModelRequest,
     TripoOrientation,
@@ -22,7 +21,6 @@ from comfy_api_nodes.apis.tripo import (
     TripoTaskType,
     TripoTextToModelRequest,
     TripoTextureModelRequest,
-    TripoTexturePrompt,
     TripoUrlReference,
 )
 from comfy_api_nodes.util import (
@@ -30,7 +28,6 @@ from comfy_api_nodes.util import (
     download_url_to_file_3d,
     poll_op,
     sync_op,
-    upload_3d_model_to_comfyapi,
     upload_images_to_comfyapi,
 )
 
@@ -541,14 +538,6 @@ class TripoTextureNode(IO.ComfyNode):
                     optional=True,
                     advanced=True,
                 ),
-                IO.String.Input(
-                    "texture_prompt",
-                    default="",
-                    multiline=True,
-                    optional=True,
-                    tooltip="Optional text guidance for texturing. Required in practice for imported "
-                    "models (Tripo: Import Model), which carry no source image to infer colors from.",
-                ),
             ],
             outputs=[
                 IO.String.Output(display_name="model_file"),  # for backward compatibility only
@@ -582,7 +571,6 @@ class TripoTextureNode(IO.ComfyNode):
         texture_seed: int | None = None,
         texture_quality: str | None = None,
         texture_alignment: str | None = None,
-        texture_prompt: str = "",
     ) -> IO.NodeOutput:
         response = await sync_op(
             cls,
@@ -595,7 +583,6 @@ class TripoTextureNode(IO.ComfyNode):
                 texture_seed=texture_seed,
                 texture_quality=texture_quality,
                 texture_alignment=texture_alignment,
-                texture_prompt=TripoTexturePrompt(text=texture_prompt.strip()) if texture_prompt.strip() else None,
             ),
         )
         return await poll_until_finished(cls, response, average_duration=80)
@@ -928,90 +915,6 @@ class TripoConversionNode(IO.ComfyNode):
         return await poll_until_finished(cls, response, average_duration=30)
 
 
-class TripoImportModelNode(IO.ComfyNode):
-    """Imports an external 3D model into Tripo, producing a MODEL_TASK_ID for post-processing nodes."""
-
-    SUPPORTED_FORMATS = ("glb", "fbx", "obj", "stl")
-
-    @classmethod
-    def define_schema(cls):
-        return IO.Schema(
-            node_id="TripoImportModelNode",
-            display_name="Tripo: Import Model",
-            category="partner/3d/Tripo",
-            description="Import an external 3D model (e.g. from Rodin, Hunyuan3D or a local file) into Tripo "
-            "to use it with Tripo's post-processing nodes: Texture, Rig, Convert. "
-            "GLB is recommended: textures survive import only when embedded in the file. "
-            "Note that texturing an imported model requires a texture prompt.",
-            inputs=[
-                IO.MultiType.Input(
-                    "model_3d",
-                    types=[IO.File3DGLB, IO.File3DFBX, IO.File3DOBJ, IO.File3DSTL, IO.File3DAny],
-                    tooltip="3D model to import (GLB / FBX / OBJ / STL, up to 150 MB). "
-                    "OBJ and STL files carry no embedded textures.",
-                ),
-            ],
-            outputs=[
-                IO.Custom("MODEL_TASK_ID").Output(display_name="model task_id"),
-            ],
-            hidden=[
-                IO.Hidden.auth_token_comfy_org,
-                IO.Hidden.api_key_comfy_org,
-                IO.Hidden.unique_id,
-            ],
-            is_api_node=True,
-            price_badge=IO.PriceBadge(
-                expr="""{"type":"text","text":"Free"}""",
-            ),
-        )
-
-    @classmethod
-    async def execute(cls, model_3d: Types.File3D) -> IO.NodeOutput:
-        file_format = (model_3d.format or "").lstrip(".").lower()
-        if file_format == "gltf":
-            raise ValueError(
-                "GLTF (.gltf) references external files and cannot be imported. Export a single-file GLB instead."
-            )
-        if file_format not in cls.SUPPORTED_FORMATS:
-            raise ValueError(
-                f"Unsupported 3D format '{file_format or 'unknown'}'. "
-                f"Tripo import supports: {', '.join(f.upper() for f in cls.SUPPORTED_FORMATS)}."
-            )
-        size = len(model_3d.get_bytes())
-        if size > 150 * 1024 * 1024:
-            raise ValueError(f"Model file is {size / (1024 * 1024):.1f} MB; Tripo import allows up to 150 MB.")
-
-        url = await upload_3d_model_to_comfyapi(cls, model_3d, file_format)
-        response = await sync_op(
-            cls,
-            endpoint=ApiEndpoint(path="/proxy/tripo/v2/openapi/import", method="POST"),
-            response_model=TripoTaskResponse,
-            data=TripoImportModelRequest(url=url, format=file_format),
-        )
-        if response.code != 0:
-            raise RuntimeError(f"Failed to import model: {response.error}")
-
-        task_id = response.data.task_id
-        response_poll = await poll_op(
-            cls,
-            poll_endpoint=ApiEndpoint(path=f"/proxy/tripo/v2/openapi/task/{task_id}"),
-            response_model=TripoTaskResponse,
-            failed_statuses=[
-                TripoTaskStatus.FAILED,
-                TripoTaskStatus.CANCELLED,
-                TripoTaskStatus.UNKNOWN,
-                TripoTaskStatus.BANNED,
-                TripoTaskStatus.EXPIRED,
-            ],
-            status_extractor=lambda x: x.data.status,
-            progress_extractor=lambda x: x.data.progress,
-            estimated_duration=10,
-        )
-        if response_poll.data.status != TripoTaskStatus.SUCCESS:
-            raise RuntimeError(f"Failed to import model: {response_poll}")
-        return IO.NodeOutput(task_id)
-
-
 def _p1_price_expr(*, geometry_credits: int, textured_credits: int, detailed_credits: int) -> str:
     return (
         "("
@@ -1389,7 +1292,6 @@ class TripoExtension(ComfyExtension):
             TripoP1TextToModelNode,
             TripoP1ImageToModelNode,
             TripoP1MultiviewToModelNode,
-            TripoImportModelNode,
             TripoTextureNode,
             TripoRefineNode,
             TripoRigNode,

comfy_extras/color_util.py

@@ -1,20 +0,0 @@
-def hex_to_rgb(value: str) -> tuple[int, int, int]:
-    h = value.lstrip("#")
-    if len(h) != 6:
-        return (255, 255, 255)
-    return (int(h[0:2], 16), int(h[2:4], 16), int(h[4:6], 16))
-
-
-def readable_color(rgb: tuple[int, int, int]) -> tuple[int, int, int]:
-    r, g, b = rgb
-    lum = 0.299 * r + 0.587 * g + 0.114 * b
-    if lum >= 130:
-        return (r, g, b)
-    t = (130 - lum) / (255 - lum)
-    return (round(r + (255 - r) * t), round(g + (255 - g) * t), round(b + (255 - b) * t))
-
-
-def normalize_palette(colors) -> list[str]:
-    if isinstance(colors, dict):
-        colors = colors.values()
-    return [c.upper() for c in colors if isinstance(c, str) and c]

comfy_extras/nodes_bounding_boxes.py

@@ -1,252 +0,0 @@
-import numpy as np
-import torch
-from PIL import Image, ImageDraw, ImageEnhance, ImageFont
-from typing_extensions import override
-
-from comfy_api.latest import ComfyExtension, io
-from comfy_extras.color_util import hex_to_rgb, normalize_palette, readable_color
-
-_PREVIEW_LONG_EDGE = 1024
-_PREVIEW_DIM = 0.25
-
-
-def pixels_to_fractions(box: dict, width: int, height: int) -> dict:
-    w = width or 1
-    h = height or 1
-    return {
-        "x": box.get("x", 0) / w,
-        "y": box.get("y", 0) / h,
-        "w": box.get("width", 0) / w,
-        "h": box.get("height", 0) / h,
-    }
-
-
-def fractions_to_pixels(box: dict, width: int, height: int) -> dict:
-    x, y = box.get("x", 0.0), box.get("y", 0.0)
-    w, h = box.get("w", 0.0), box.get("h", 0.0)
-    if w < 0:
-        x, w = x + w, -w
-    if h < 0:
-        y, h = y + h, -h
-    return {
-        "x": round(x * width),
-        "y": round(y * height),
-        "width": round(w * width),
-        "height": round(h * height),
-    }
-
-
-def fractions_to_bbox_frame(boxes: list, width: int, height: int) -> list:
-    pixels = [
-        fractions_to_pixels(box, width, height)
-        for box in boxes
-        if isinstance(box, dict)
-    ]
-    return [pixels] if pixels else []
-
-
-def _font(size: int):
-    try:
-        return ImageFont.load_default(size)
-    except Exception:
-        return ImageFont.load_default()
-
-
-def _wrap(draw, text: str, font, max_w: float) -> list[str]:
-    lines = []
-    for para in text.split("\n"):
-        line = ""
-        for word in para.split():
-            test = word if not line else line + " " + word
-            if line and draw.textlength(test, font=font) > max_w:
-                lines.append(line)
-                line = word
-            else:
-                line = test
-        lines.append(line)
-    return lines
-
-
-def _bg_from_image(image) -> Image.Image | None:
-    if image is None:
-        return None
-    try:
-        arr = (image[0].detach().cpu().numpy() * 255).clip(0, 255).astype(np.uint8)
-        return Image.fromarray(arr)
-    except Exception:
-        return None
-
-
-def render_preview(regions, width, height, bg=None):
-    if bg is not None:
-        iw, ih = bg.size
-        long_edge = max(iw, ih) or 1
-        scale = min(1.0, _PREVIEW_LONG_EDGE / long_edge)
-        rw, rh = max(1, round(iw * scale)), max(1, round(ih * scale))
-        base = bg.convert("RGB").resize((rw, rh), Image.LANCZOS)
-        base = ImageEnhance.Brightness(base).enhance(_PREVIEW_DIM)
-        img = base.convert("RGBA")
-    else:
-        long_edge = max(width, height) or 1
-        scale = min(1.0, _PREVIEW_LONG_EDGE / long_edge)
-        rw, rh = max(1, round(width * scale)), max(1, round(height * scale))
-        grey = round(_PREVIEW_DIM * 128)
-        img = Image.new("RGBA", (rw, rh), (grey, grey, grey, 255))
-
-    overlay = Image.new("RGBA", (rw, rh), (0, 0, 0, 0))
-    draw = ImageDraw.Draw(overlay)
-    fs = max(10, round(rh / 64))
-    font = _font(fs)
-    tag_font = _font(max(9, fs - 2))
-    line_h = fs + 2
-
-    for i, region in enumerate(regions):
-        if not isinstance(region, dict):
-            continue
-        palette = [c for c in (region.get("palette") or []) if c]
-        r, g, b = hex_to_rgb(palette[0]) if palette else (140, 140, 140)
-        x1 = max(0, min(rw, round(region.get("x", 0) * rw)))
-        y1 = max(0, min(rh, round(region.get("y", 0) * rh)))
-        x2 = max(0, min(rw, round((region.get("x", 0) + region.get("w", 0)) * rw)))
-        y2 = max(0, min(rh, round((region.get("y", 0) + region.get("h", 0)) * rh)))
-        if x2 < x1:
-            x1, x2 = x2, x1
-        if y2 < y1:
-            y1, y2 = y2, y1
-
-        draw.rectangle([x1, y1, x2, y2], outline=(r, g, b, 255), width=2)
-
-        swatches = palette[:5]
-        if swatches and (x2 - x1) > 2:
-            sh = max(5, fs // 2)
-            seg = (x2 - x1) / len(swatches)
-            for p, hexc in enumerate(swatches):
-                sx = x1 + round(p * seg)
-                draw.rectangle([sx, y1, x1 + round((p + 1) * seg), y1 + sh], fill=hex_to_rgb(hexc))
-
-        etype = "text" if region.get("type") == "text" else "obj"
-        tag = str(i + 1).zfill(2)
-        tw = draw.textlength(tag, font=tag_font)
-        draw.rectangle([x1, y1, x1 + tw + 6, y1 + fs + 2], fill=(r, g, b, 255))
-        tag_fill = (0, 0, 0, 255) if (0.299 * r + 0.587 * g + 0.114 * b) > 140 else (255, 255, 255, 255)
-        draw.text((x1 + 3, y1 + 1), tag, fill=tag_fill, font=tag_font)
-
-        body = region.get("desc", "") or ""
-        if etype == "text" and region.get("text"):
-            body = '"%s"%s' % (region["text"], " — " + body if body else "")
-        if body and (x2 - x1) > 8:
-            ty = y1 + fs + 5
-            for line in _wrap(draw, body, font, x2 - x1 - 8):
-                if ty > y2:
-                    break
-                draw.text((x1 + 4, ty), line, fill=readable_color((r, g, b)) + (255,), font=font)
-                ty += line_h
-
-    composed = Image.alpha_composite(img, overlay).convert("RGB")
-    arr = np.asarray(composed, dtype=np.float32) / 255.0
-    return torch.from_numpy(arr).unsqueeze(0)
-
-
-def boxes_to_regions(boxes, width: int, height: int) -> list:
-    regions: list = []
-    if not isinstance(boxes, list):
-        return regions
-    for box in boxes:
-        if not isinstance(box, dict):
-            continue
-        meta = box.get("metadata")
-        meta = meta if isinstance(meta, dict) else {}
-        regions.append({
-            **pixels_to_fractions(box, width, height),
-            "type": meta.get("type", "obj"),
-            "text": meta.get("text", ""),
-            "desc": meta.get("desc", ""),
-            "palette": meta.get("palette", []),
-        })
-    return regions
-
-
-def _norm_bbox(region: dict) -> list[int]:
-    def grid(value: float) -> int:
-        return max(0, min(1000, round(value * 1000)))
-
-    x, y = region.get("x", 0.0), region.get("y", 0.0)
-    w, h = region.get("w", 0.0), region.get("h", 0.0)
-    ymin, xmin, ymax, xmax = grid(y), grid(x), grid(y + h), grid(x + w)
-    if ymin > ymax:
-        ymin, ymax = ymax, ymin
-    if xmin > xmax:
-        xmin, xmax = xmax, xmin
-    return [ymin, xmin, ymax, xmax]
-
-
-def build_elements(regions: list) -> list:
-    elements = []
-    for region in regions:
-        if not isinstance(region, dict):
-            continue
-        etype = "text" if region.get("type") == "text" else "obj"
-        element = {"type": etype}
-        element["bbox"] = _norm_bbox(region)
-        if etype == "text":
-            element["text"] = region.get("text", "")
-        element["desc"] = region.get("desc", "")
-        palette = normalize_palette(region.get("palette", []))
-        if palette:
-            element["color_palette"] = palette[:5]
-        elements.append(element)
-    return elements
-
-
-class CreateBoundingBoxes(io.ComfyNode):
-    @classmethod
-    def define_schema(cls):
-        editor_state = io.BoundingBoxes.Input(
-            "editor_state",
-            tooltip="Draw regions and set each region's type/text/desc/palette.",
-        )
-        return io.Schema(
-            node_id="CreateBoundingBoxes",
-            display_name="Create Bounding Boxes",
-            category="utilities",
-            description="Draw regions over a reference image. Outputs Ideogram caption elements, pixel-space bounding boxes, and a rendered preview.",
-            inputs=[
-                io.Image.Input(
-                    "background",
-                    optional=True,
-                    tooltip="Optional reference image shown behind the canvas and preview.",
-                ),
-                io.Int.Input("width", default=1024, min=64, max=16384, step=16,
-                             tooltip="Canvas aspect width and the pixel grid for the bbox output."),
-                io.Int.Input("height", default=1024, min=64, max=16384, step=16,
-                             tooltip="Canvas aspect height and the pixel grid for the bbox output."),
-                editor_state,
-            ],
-            outputs=[
-                io.Image.Output(display_name="preview"),
-                io.BoundingBox.Output(display_name="bboxes"),
-                io.ComfyList.Output(display_name="elements"),
-            ],
-            is_experimental=True,
-        )
-
-    @classmethod
-    def execute(cls, width, height, editor_state=None, background=None) -> io.NodeOutput:
-        regions = boxes_to_regions(editor_state, width, height)
-        preview = render_preview(regions, width, height, _bg_from_image(background))
-        return io.NodeOutput(
-            preview,
-            fractions_to_bbox_frame(regions, width, height),
-            build_elements(regions),
-            ui={"dims": [width, height]},
-        )
-
-
-class BoundingBoxesExtension(ComfyExtension):
-    @override
-    async def get_node_list(self) -> list[type[io.ComfyNode]]:
-        return [CreateBoundingBoxes]
-
-
-async def comfy_entrypoint() -> BoundingBoxesExtension:
-    return BoundingBoxesExtension()

comfy_extras/nodes_json_prompt.py

@@ -1,76 +0,0 @@
-from typing_extensions import override
-
-from comfy_api.latest import ComfyExtension, io
-from comfy_extras.color_util import normalize_palette
-
-
-class BuildJsonPromptIdeogram(io.ComfyNode):
-    @classmethod
-    def define_schema(cls):
-        color_palette = io.Colors.Input(
-            "color_palette",
-            tooltip="Style color palette.",
-        )
-        return io.Schema(
-            node_id="BuildJsonPromptIdeogram",
-            display_name="Build JSON Prompt (Ideogram)",
-            category="image/ideogram",
-            description="Assemble the Ideogram 4 caption from Create Bounding Boxes elements plus the background and style fields.",
-            inputs=[
-                io.ComfyList.Input("element", tooltip="Caption elements from Create Bounding Boxes."),
-                io.String.Input("high_level_description", multiline=True, default="",
-                                tooltip="Optional one-line overview of the whole image (blank = omitted)."),
-                io.String.Input("background", multiline=True, default="",
-                                tooltip="Scene background description."),
-                io.DynamicCombo.Input("style", options=[
-                    io.DynamicCombo.Option("none", []),
-                    io.DynamicCombo.Option("photo", [io.String.Input("photo", default="")]),
-                    io.DynamicCombo.Option("art_style", [io.String.Input("art_style", default="")]),
-                ]),
-                io.String.Input("aesthetics", default="", tooltip="Style descriptor (blank = omitted once a style is chosen)."),
-                io.String.Input("lighting", default="", tooltip="Style descriptor (blank = omitted once a style is chosen)."),
-                io.String.Input("medium", default="", tooltip="Style descriptor (blank = omitted once a style is chosen)."),
-                color_palette,
-            ],
-            outputs=[io.ComfyDict.Output(display_name="prompt")],
-            is_experimental=True,
-        )
-
-    @classmethod
-    def execute(cls, element, style, high_level_description="", background="",
-                aesthetics="", lighting="", medium="", color_palette=None) -> io.NodeOutput:
-        elements = element if isinstance(element, list) else []
-        kind = style.get("style", "none") if isinstance(style, dict) else "none"
-        photo = style.get("photo", "") if isinstance(style, dict) else ""
-        art_style = style.get("art_style", "") if isinstance(style, dict) else ""
-        palette = normalize_palette(color_palette or [])
-
-        caption: dict = {}
-        if high_level_description.strip():
-            caption["high_level_description"] = high_level_description
-        if kind != "none":
-            style_desc: dict = {"aesthetics": aesthetics, "lighting": lighting}
-            if kind == "photo":
-                style_desc["photo"] = photo
-                style_desc["medium"] = medium
-            else:
-                style_desc["medium"] = medium
-                style_desc["art_style"] = art_style
-            if palette:
-                style_desc["color_palette"] = palette
-            caption["style_description"] = style_desc
-        caption["compositional_deconstruction"] = {
-            "background": background,
-            "elements": elements,
-        }
-        return io.NodeOutput(caption)
-
-
-class JsonPromptExtension(ComfyExtension):
-    @override
-    async def get_node_list(self) -> list[type[io.ComfyNode]]:
-        return [BuildJsonPromptIdeogram]
-
-
-async def comfy_entrypoint() -> JsonPromptExtension:
-    return JsonPromptExtension()

comfy_extras/nodes_string.py

@@ -440,30 +440,6 @@ class JsonExtractString(io.ComfyNode):
         except (json.JSONDecodeError, TypeError):
             return io.NodeOutput("")
 
-
-class DictToJsonString(io.ComfyNode):
-    @classmethod
-    def define_schema(cls):
-        return io.Schema(
-            node_id="DictToJsonString",
-            display_name="Dict to JSON String",
-            category="text",
-            search_aliases=["json", "dict to json", "stringify", "serialize", "dict to string"],
-            inputs=[
-                io.ComfyDict.Input("value"),
-                io.Int.Input("indent", default=2, min=0, max=8,
-                             tooltip="Spaces per indent level. 0 produces compact single-line JSON."),
-            ],
-            outputs=[
-                io.String.Output(),
-            ],
-        )
-
-    @classmethod
-    def execute(cls, value, indent=2):
-        return io.NodeOutput(json.dumps(value, ensure_ascii=False, indent=indent or None))
-
-
 class StringExtension(ComfyExtension):
     @override
     async def get_node_list(self) -> list[type[io.ComfyNode]]:
@@ -481,7 +457,6 @@ class StringExtension(ComfyExtension):
             RegexExtract,
             RegexReplace,
             JsonExtractString,
-            DictToJsonString,
         ]
 
 async def comfy_entrypoint() -> StringExtension:

comfy_extras/nodes_video.py

@@ -134,17 +134,6 @@ class CreateVideo(io.ComfyNode):
                 io.Image.Input("images", tooltip="The images to create a video from."),
                 io.Float.Input("fps", default=30.0, min=1.0, max=120.0, step=1.0),
                 io.Audio.Input("audio", optional=True, tooltip="The audio to add to the video."),
-                io.Int.Input(
-                    "bit_depth",
-                    min=8,
-                    max=10,
-                    default=8,
-                    step=2,
-                    tooltip="Bit depth of the created video. 10-bit keeps smoother gradients with less"
-                    " banding, but some players and downstream nodes may not support it.",
-                    optional=True,
-                    display_mode=io.NumberDisplay.number,
-                ),
             ],
             outputs=[
                 io.Video.Output(),
@@ -152,14 +141,9 @@ class CreateVideo(io.ComfyNode):
         )
 
     @classmethod
-    def execute(
-        cls, images: Input.Image, fps: float, audio: Optional[Input.Audio] = None, bit_depth: int = 8,
-    ) -> io.NodeOutput:
+    def execute(cls, images: Input.Image, fps: float, audio: Optional[Input.Audio] = None) -> io.NodeOutput:
         return io.NodeOutput(
-            InputImpl.VideoFromComponents(
-                Types.VideoComponents(images=images, audio=audio, frame_rate=Fraction(fps)),
-                bit_depth=bit_depth,
-            )
+            InputImpl.VideoFromComponents(Types.VideoComponents(images=images, audio=audio, frame_rate=Fraction(fps)))
         )
 
 class GetVideoComponents(io.ComfyNode):
@@ -170,7 +154,7 @@ class GetVideoComponents(io.ComfyNode):
             search_aliases=["extract frames", "split video", "video to images", "demux"],
             display_name="Get Video Components",
             category="video",
-            description="Extracts all components from a video: frames, audio, framerate, and bit depth.",
+            description="Extracts all components from a video: frames, audio, and framerate.",
             inputs=[
                 io.Video.Input("video", tooltip="The video to extract components from."),
             ],
@@ -178,14 +162,13 @@ class GetVideoComponents(io.ComfyNode):
                 io.Image.Output(display_name="images"),
                 io.Audio.Output(display_name="audio"),
                 io.Float.Output(display_name="fps"),
-                io.Int.Output(display_name="bit_depth"),
             ],
         )
 
     @classmethod
     def execute(cls, video: Input.Video) -> io.NodeOutput:
         components = video.get_components()
-        return io.NodeOutput(components.images, components.audio, float(components.frame_rate), video.get_bit_depth())
+        return io.NodeOutput(components.images, components.audio, float(components.frame_rate))
 
 
 class LoadVideo(io.ComfyNode):

manager_requirements.txt

@@ -1 +1 @@
-comfyui_manager==4.2.2
+comfyui_manager==4.2.1

nodes.py

@@ -2363,8 +2363,6 @@ async def init_builtin_extra_nodes():
         "nodes_images.py",
         "nodes_video_model.py",
         "nodes_ideogram4.py",
-        "nodes_bounding_boxes.py",
-        "nodes_json_prompt.py",
         "nodes_train.py",
         "nodes_dataset.py",
         "nodes_sag.py",

requirements.txt

@@ -1,6 +1,6 @@
 comfyui-frontend-package==1.45.15
 comfyui-workflow-templates==0.9.98
-comfyui-embedded-docs==0.5.4
+comfyui-embedded-docs==0.5.3
 torch
 torchsde
 torchvision

server.py

@@ -27,7 +27,6 @@ import logging
 
 import mimetypes
 from comfy.cli_args import args
-from comfy.deploy_environment import get_deploy_environment
 import comfy.utils
 import comfy.model_management
 from comfy_api import feature_flags
@@ -691,7 +690,6 @@ class PromptServer():
                     "python_version": sys.version,
                     "pytorch_version": comfy.model_management.torch_version,
                     "embedded_python": os.path.split(os.path.split(sys.executable)[0])[1] == "python_embeded",
-                    "deploy_environment": get_deploy_environment(),
                     "argv": sys.argv
                 },
                 "devices": device_entries

tests-unit/comfy_api_test/video_bit_depth_test.py

@@ -1,93 +0,0 @@
-import pytest
-import torch
-import av
-import numpy as np
-from fractions import Fraction
-from comfy_api.latest._input_impl.video_types import VideoFromFile, VideoFromComponents
-from comfy_api.latest._util.video_types import VideoComponents
-
-
-@pytest.fixture(scope="module")
-def gradient_components():
-    """Narrow horizontal ramp (0.25..0.30) that needs more than 8 bits to stay smooth"""
-    width, height, frames = 64, 64, 3
-    ramp = torch.linspace(0.25, 0.30, width).view(1, 1, width, 1).expand(frames, height, width, 3)
-    return VideoComponents(images=ramp.contiguous(), frame_rate=Fraction(30))
-
-
-@pytest.fixture(scope="module")
-def src8(gradient_components, tmp_path_factory):
-    """8-bit h264 mp4 (Create Video default)"""
-    path = str(tmp_path_factory.mktemp("video") / "src8.mp4")
-    VideoFromComponents(gradient_components).save_to(path)
-    return path
-
-
-@pytest.fixture(scope="module")
-def src10(gradient_components, tmp_path_factory):
-    """10-bit h264 mp4 (Create Video with bit_depth=10)"""
-    path = str(tmp_path_factory.mktemp("video") / "src10.mp4")
-    VideoFromComponents(gradient_components, bit_depth=10).save_to(path)
-    return path
-
-
-def probe(path):
-    """(codec, pix_fmt, bit_depth) of the first video stream"""
-    with av.open(path) as container:
-        stream = container.streams.video[0]
-        return (stream.codec.name, stream.format.name, max(c.bits for c in stream.format.components))
-
-
-def decoded_levels(path):
-    """Unique tonal levels in the first decoded frame (banding measure)"""
-    with av.open(path) as container:
-        frame = next(container.decode(container.streams.video[0]))
-        return len(np.unique(frame.to_ndarray(format="gbrpf32le")[..., 0]))
-
-
-def video_packet_bytes(path):
-    """Raw video packet payloads; identical to the source's only for a true remux"""
-    with av.open(path) as container:
-        return [bytes(p) for p in container.demux(container.streams.video[0]) if p.size]
-
-
-def test_create_video_bit_depth(src8, src10):
-    """Create Video's bit_depth picks the encoded depth (default 8-bit); 10-bit reduces banding"""
-    assert probe(src8) == ("h264", "yuv420p", 8)
-    assert probe(src10) == ("h264", "yuv420p10le", 10)
-    assert decoded_levels(src10) > 2 * decoded_levels(src8)
-
-
-def test_save_auto_keeps_source_depth(src8, src10, tmp_path):
-    """Save Video (no bit_depth = auto) stream-copies the source, preserving its depth byte-for-byte"""
-    for name, src in [("p8", src8), ("p10", src10)]:
-        path = str(tmp_path / f"{name}.mp4")
-        VideoFromFile(src).save_to(path)
-        assert probe(path) == probe(src)
-        assert video_packet_bytes(path) == video_packet_bytes(src)
-
-
-def test_save_explicit_depth_reencodes(src8, src10, tmp_path):
-    """An explicit bit_depth different from the source forces a re-encode to that depth"""
-    down = str(tmp_path / "down8.mp4")
-    VideoFromFile(src10).save_to(down, bit_depth=8)
-    assert probe(down) == ("h264", "yuv420p", 8)
-
-    up = str(tmp_path / "up10.mp4")
-    VideoFromFile(src8).save_to(up, bit_depth=10)
-    assert probe(up) == ("h264", "yuv420p10le", 10)
-
-
-def test_trim_keeps_source_depth(src10, tmp_path):
-    """Video Slice re-encodes (trim) but preserves the source's 10-bit depth"""
-    path = str(tmp_path / "trim.mp4")
-    VideoFromFile(src10).as_trimmed(start_time=0, duration=1 / 30, strict_duration=False).save_to(path)
-    assert probe(path) == ("h264", "yuv420p10le", 10)
-
-
-def test_get_bit_depth(gradient_components, src8, src10):
-    """get_bit_depth reports a video's depth (backs the Get Video Components output)"""
-    assert VideoFromFile(src8).get_bit_depth() == 8
-    assert VideoFromFile(src10).get_bit_depth() == 10
-    assert VideoFromComponents(gradient_components, bit_depth=10).get_bit_depth() == 10
-    assert VideoFromComponents(gradient_components).get_bit_depth() == 8