feat: add timestamp to default filename_prefix for cache-busting

Change default filename_prefix on all previewable save nodes (image, video,
audio, 3D, SVG) from 'ComfyUI' to 'ComfyUI_%year%%month%%day%-%hour%%minute%%second%'.

This leverages the existing compute_vars template system in
get_save_image_path — zero new backend code needed. Each output gets a
unique filename per second, preventing browser cache from showing stale
previews when files are overwritten.

Users can customize or remove the template from the node widget.
Existing workflows retain their saved prefix value (only new nodes
get the new default). Custom nodes are unaffected — they define their
own defaults independently.

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -16,7 +16,7 @@ body:
 
         ## Very Important
 
-        Please make sure that you post ALL your ComfyUI logs in the bug report. A bug report without logs will likely be ignored.
+        Please make sure that you post ALL your ComfyUI logs in the bug report **even if there is no crash**. Just paste everything. The startup log (everything before "To see the GUI go to: ...") contains critical information to developers trying to help. For a performance issue or crash, paste everything from "got prompt" to the end, including the crash. More is better - always. A bug report without logs will likely be ignored.
   - type: checkboxes
     id: custom-nodes-test
     attributes:

README.md

@@ -189,8 +189,6 @@ The portable above currently comes with python 3.13 and pytorch cuda 13.0. Updat
 
 [Experimental portable for AMD GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_amd.7z)
 
-[Portable with pytorch cuda 12.8 and python 3.12](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia_cu128.7z).
-
 [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?

comfy/ldm/modules/diffusionmodules/openaimodel.py

@@ -18,6 +18,8 @@ import comfy.patcher_extension
 import comfy.ops
 ops = comfy.ops.disable_weight_init
 
+from ..sdpose import HeatmapHead
+
 class TimestepBlock(nn.Module):
     """
     Any module where forward() takes timestep embeddings as a second argument.
@@ -441,6 +443,7 @@ class UNetModel(nn.Module):
         disable_temporal_crossattention=False,
         max_ddpm_temb_period=10000,
         attn_precision=None,
+        heatmap_head=False,
         device=None,
         operations=ops,
     ):
@@ -827,6 +830,9 @@ class UNetModel(nn.Module):
             #nn.LogSoftmax(dim=1)  # change to cross_entropy and produce non-normalized logits
         )
 
+        if heatmap_head:
+            self.heatmap_head = HeatmapHead(device=device, dtype=self.dtype, operations=operations)
+
     def forward(self, x, timesteps=None, context=None, y=None, control=None, transformer_options={}, **kwargs):
         return comfy.patcher_extension.WrapperExecutor.new_class_executor(
             self._forward,

comfy/ldm/modules/sdpose.py

@@ -0,0 +1,130 @@
+import torch
+import numpy as np
+from scipy.ndimage import gaussian_filter
+
+class HeatmapHead(torch.nn.Module):
+    def __init__(
+            self,
+            in_channels=640,
+            out_channels=133,
+            input_size=(768, 1024),
+            heatmap_scale=4,
+            deconv_out_channels=(640,),
+            deconv_kernel_sizes=(4,),
+            conv_out_channels=(640,),
+            conv_kernel_sizes=(1,),
+            final_layer_kernel_size=1,
+            device=None, dtype=None, operations=None
+        ):
+        super().__init__()
+
+        self.heatmap_size = (input_size[0] // heatmap_scale, input_size[1] // heatmap_scale)
+        self.scale_factor = ((np.array(input_size) - 1) / (np.array(self.heatmap_size) - 1)).astype(np.float32)
+
+        # Deconv layers
+        if deconv_out_channels:
+            deconv_layers = []
+            for out_ch, kernel_size in zip(deconv_out_channels, deconv_kernel_sizes):
+                if kernel_size == 4:
+                    padding, output_padding = 1, 0
+                elif kernel_size == 3:
+                    padding, output_padding = 1, 1
+                elif kernel_size == 2:
+                    padding, output_padding = 0, 0
+                else:
+                    raise ValueError(f'Unsupported kernel size {kernel_size}')
+
+                deconv_layers.extend([
+                    operations.ConvTranspose2d(in_channels, out_ch, kernel_size,
+                                     stride=2, padding=padding, output_padding=output_padding, bias=False, device=device, dtype=dtype),
+                    torch.nn.InstanceNorm2d(out_ch, device=device, dtype=dtype),
+                    torch.nn.SiLU(inplace=True)
+                ])
+                in_channels = out_ch
+            self.deconv_layers = torch.nn.Sequential(*deconv_layers)
+        else:
+            self.deconv_layers = torch.nn.Identity()
+
+        # Conv layers
+        if conv_out_channels:
+            conv_layers = []
+            for out_ch, kernel_size in zip(conv_out_channels, conv_kernel_sizes):
+                padding = (kernel_size - 1) // 2
+                conv_layers.extend([
+                    operations.Conv2d(in_channels, out_ch, kernel_size,
+                            stride=1, padding=padding, device=device, dtype=dtype),
+                    torch.nn.InstanceNorm2d(out_ch, device=device, dtype=dtype),
+                    torch.nn.SiLU(inplace=True)
+                ])
+                in_channels = out_ch
+            self.conv_layers = torch.nn.Sequential(*conv_layers)
+        else:
+            self.conv_layers = torch.nn.Identity()
+
+        self.final_layer = operations.Conv2d(in_channels, out_channels, kernel_size=final_layer_kernel_size, padding=final_layer_kernel_size // 2, device=device, dtype=dtype)
+
+    def forward(self, x): # Decode heatmaps to keypoints
+        heatmaps = self.final_layer(self.conv_layers(self.deconv_layers(x)))
+        heatmaps_np = heatmaps.float().cpu().numpy()  # (B, K, H, W)
+        B, K, H, W = heatmaps_np.shape
+
+        batch_keypoints = []
+        batch_scores = []
+
+        for b in range(B):
+            hm = heatmaps_np[b].copy()  # (K, H, W)
+
+            # --- vectorised argmax ---
+            flat = hm.reshape(K, -1)
+            idx = np.argmax(flat, axis=1)
+            scores = flat[np.arange(K), idx].copy()
+            y_locs, x_locs = np.unravel_index(idx, (H, W))
+            keypoints = np.stack([x_locs, y_locs], axis=-1).astype(np.float32)  # (K, 2) in heatmap space
+            invalid = scores <= 0.
+            keypoints[invalid] = -1
+
+            # --- DARK sub-pixel refinement (UDP) ---
+            # 1. Gaussian blur with max-preserving normalisation
+            border = 5  # (kernel-1)//2 for kernel=11
+            for k in range(K):
+                origin_max = np.max(hm[k])
+                dr = np.zeros((H + 2 * border, W + 2 * border), dtype=np.float32)
+                dr[border:-border, border:-border] = hm[k].copy()
+                dr = gaussian_filter(dr, sigma=2.0)
+                hm[k] = dr[border:-border, border:-border].copy()
+                cur_max = np.max(hm[k])
+                if cur_max > 0:
+                    hm[k] *= origin_max / cur_max
+            # 2. Log-space for Taylor expansion
+            np.clip(hm, 1e-3, 50., hm)
+            np.log(hm, hm)
+            # 3. Hessian-based Newton step
+            hm_pad = np.pad(hm, ((0, 0), (1, 1), (1, 1)), mode='edge').flatten()
+            index = keypoints[:, 0] + 1 + (keypoints[:, 1] + 1) * (W + 2)
+            index += (W + 2) * (H + 2) * np.arange(0, K)
+            index = index.astype(int).reshape(-1, 1)
+            i_       = hm_pad[index]
+            ix1      = hm_pad[index + 1]
+            iy1      = hm_pad[index + W + 2]
+            ix1y1    = hm_pad[index + W + 3]
+            ix1_y1_  = hm_pad[index - W - 3]
+            ix1_     = hm_pad[index - 1]
+            iy1_     = hm_pad[index - 2 - W]
+            dx = 0.5 * (ix1 - ix1_)
+            dy = 0.5 * (iy1 - iy1_)
+            derivative = np.concatenate([dx, dy], axis=1).reshape(K, 2, 1)
+            dxx = ix1  - 2 * i_ + ix1_
+            dyy = iy1  - 2 * i_ + iy1_
+            dxy = 0.5 * (ix1y1 - ix1 - iy1 + i_ + i_ - ix1_ - iy1_ + ix1_y1_)
+            hessian = np.concatenate([dxx, dxy, dxy, dyy], axis=1).reshape(K, 2, 2)
+            hessian = np.linalg.inv(hessian + np.finfo(np.float32).eps * np.eye(2))
+            keypoints -= np.einsum('imn,ink->imk', hessian, derivative).squeeze(axis=-1)
+
+            # --- restore to input image space ---
+            keypoints = keypoints * self.scale_factor
+            keypoints[invalid] = -1
+
+            batch_keypoints.append(keypoints)
+            batch_scores.append(scores)
+
+        return batch_keypoints, batch_scores

comfy/ldm/wan/vae.py

@@ -485,7 +485,7 @@ class WanVAE(nn.Module):
         iter_ = 1 + (t - 1) // 4
         feat_map = None
         if iter_ > 1:
-            feat_map = [None] * count_conv3d(self.decoder)
+            feat_map = [None] * count_conv3d(self.encoder)
         ## 对encode输入的x，按时间拆分为1、4、4、4....
         for i in range(iter_):
             conv_idx = [0]

comfy/lora.py

@@ -337,6 +337,7 @@ def model_lora_keys_unet(model, key_map={}):
             if k.startswith("diffusion_model.decoder.") and k.endswith(".weight"):
                 key_lora = k[len("diffusion_model.decoder."):-len(".weight")]
                 key_map["base_model.model.{}".format(key_lora)] = k  # Official base model loras
+                key_map["lycoris_{}".format(key_lora.replace(".", "_"))] = k  # LyCORIS/LoKR format
 
     return key_map
 

comfy/model_base.py

@@ -925,6 +925,25 @@ class Flux(BaseModel):
             out['ref_latents'] = list([1, 16, sum(map(lambda a: math.prod(a.size()[2:]), ref_latents))])
         return out
 
+class LongCatImage(Flux):
+    def _apply_model(self, x, t, c_concat=None, c_crossattn=None, control=None, transformer_options={}, **kwargs):
+        transformer_options = transformer_options.copy()
+        rope_opts = transformer_options.get("rope_options", {})
+        rope_opts = dict(rope_opts)
+        rope_opts.setdefault("shift_t", 1.0)
+        rope_opts.setdefault("shift_y", 512.0)
+        rope_opts.setdefault("shift_x", 512.0)
+        transformer_options["rope_options"] = rope_opts
+        return super()._apply_model(x, t, c_concat, c_crossattn, control, transformer_options, **kwargs)
+
+    def encode_adm(self, **kwargs):
+        return None
+
+    def extra_conds(self, **kwargs):
+        out = super().extra_conds(**kwargs)
+        out.pop('guidance', None)
+        return out
+
 class Flux2(Flux):
     def extra_conds(self, **kwargs):
         out = super().extra_conds(**kwargs)

comfy/model_detection.py

@@ -279,6 +279,8 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
             dit_config["txt_norm"] = any_suffix_in(state_dict_keys, key_prefix, 'txt_norm.', ["weight", "scale"])
             if dit_config["yak_mlp"] and dit_config["txt_norm"]:  # Ovis model
                 dit_config["txt_ids_dims"] = [1, 2]
+            if dit_config.get("context_in_dim") == 3584 and dit_config["vec_in_dim"] is None:  # LongCat-Image
+                dit_config["txt_ids_dims"] = [1, 2]
 
         return dit_config
 
@@ -795,6 +797,10 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
         unet_config["use_temporal_resblock"] = False
         unet_config["use_temporal_attention"] = False
 
+    heatmap_key = '{}heatmap_head.conv_layers.0.weight'.format(key_prefix)
+    if heatmap_key in state_dict_keys:
+        unet_config["heatmap_head"] = True
+
     return unet_config
 
 def model_config_from_unet_config(unet_config, state_dict=None):
@@ -1015,7 +1021,7 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
 
     LotusD = {'use_checkpoint': False, 'image_size': 32, 'out_channels': 4, 'use_spatial_transformer': True, 'legacy': False, 'adm_in_channels': 4,
             'dtype': dtype, 'in_channels': 4, 'model_channels': 320, 'num_res_blocks': [2, 2, 2, 2], 'transformer_depth': [1, 1, 1, 1, 1, 1, 0, 0],
-            'channel_mult': [1, 2, 4, 4], 'transformer_depth_middle': 1, 'use_linear_in_transformer': True, 'context_dim': 1024, 'num_heads': 8,
+            'channel_mult': [1, 2, 4, 4], 'transformer_depth_middle': 1, 'use_linear_in_transformer': True, 'context_dim': 1024, 'num_head_channels': 64,
             'transformer_depth_output': [1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
             'use_temporal_attention': False, 'use_temporal_resblock': False}
 

comfy/model_management.py

@@ -350,7 +350,7 @@ AMD_ENABLE_MIOPEN_ENV = 'COMFYUI_ENABLE_MIOPEN'
 
 try:
     if is_amd():
-        arch = torch.cuda.get_device_properties(get_torch_device()).gcnArchName
+        arch = torch.cuda.get_device_properties(get_torch_device()).gcnArchName.split(':')[0]
         if not (any((a in arch) for a in AMD_RDNA2_AND_OLDER_ARCH)):
             if os.getenv(AMD_ENABLE_MIOPEN_ENV) != '1':
                 torch.backends.cudnn.enabled = False  # Seems to improve things a lot on AMD
@@ -378,7 +378,7 @@ try:
         if args.use_split_cross_attention == False and args.use_quad_cross_attention == False:
             if aotriton_supported(arch):  # AMD efficient attention implementation depends on aotriton.
                 if torch_version_numeric >= (2, 7):  # works on 2.6 but doesn't actually seem to improve much
-                    if any((a in arch) for a in ["gfx90a", "gfx942", "gfx1100", "gfx1101", "gfx1151"]):  # TODO: more arches, TODO: gfx950
+                    if any((a in arch) for a in ["gfx90a", "gfx942", "gfx950", "gfx1100", "gfx1101", "gfx1151"]):  # TODO: more arches, TODO: gfx950
                         ENABLE_PYTORCH_ATTENTION = True
                 if rocm_version >= (7, 0):
                    if any((a in arch) for a in ["gfx1200", "gfx1201"]):

comfy/ops.py

@@ -167,17 +167,15 @@ def cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compu
             x = to_dequant(x, dtype)
         if not resident and lowvram_fn is not None:
             x = to_dequant(x, dtype if compute_dtype is None else compute_dtype)
-            #FIXME: this is not accurate, we need to be sensitive to the compute dtype
             x = lowvram_fn(x)
-            if (isinstance(orig, QuantizedTensor) and
-                (want_requant and len(fns) == 0 or update_weight)):
+            if (want_requant and len(fns) == 0 or update_weight):
                 seed = comfy.utils.string_to_seed(s.seed_key)
-                y = QuantizedTensor.from_float(x, s.layout_type, scale="recalculate", stochastic_rounding=seed)
-                if want_requant and len(fns) == 0:
-                    #The layer actually wants our freshly saved QT
-                    x = y
-            elif update_weight:
-                y = comfy.float.stochastic_rounding(x, orig.dtype, seed = comfy.utils.string_to_seed(s.seed_key))
+                if isinstance(orig, QuantizedTensor):
+                    y = QuantizedTensor.from_float(x, s.layout_type, scale="recalculate", stochastic_rounding=seed)
+                else:
+                    y = comfy.float.stochastic_rounding(x, orig.dtype, seed=seed)
+            if want_requant and len(fns) == 0:
+                x = y
             if update_weight:
                 orig.copy_(y)
         for f in fns:
@@ -617,7 +615,8 @@ def fp8_linear(self, input):
 
     if input.ndim != 2:
         return None
-    w, bias, offload_stream = cast_bias_weight(self, input, dtype=dtype, bias_dtype=input_dtype, offloadable=True)
+    lora_compute_dtype=comfy.model_management.lora_compute_dtype(input.device)
+    w, bias, offload_stream = cast_bias_weight(self, input, dtype=dtype, bias_dtype=input_dtype, offloadable=True, compute_dtype=lora_compute_dtype, want_requant=True)
     scale_weight = torch.ones((), device=input.device, dtype=torch.float32)
 
     scale_input = torch.ones((), device=input.device, dtype=torch.float32)

comfy/sd.py

@@ -60,6 +60,7 @@ import comfy.text_encoders.jina_clip_2
 import comfy.text_encoders.newbie
 import comfy.text_encoders.anima
 import comfy.text_encoders.ace15
+import comfy.text_encoders.longcat_image
 
 import comfy.model_patcher
 import comfy.lora
@@ -1160,6 +1161,7 @@ class CLIPType(Enum):
     KANDINSKY5_IMAGE = 23
     NEWBIE = 24
     FLUX2 = 25
+    LONGCAT_IMAGE = 26
 
 
 def load_clip(ckpt_paths, embedding_directory=None, clip_type=CLIPType.STABLE_DIFFUSION, model_options={}):
@@ -1372,6 +1374,9 @@ def load_text_encoder_state_dicts(state_dicts=[], embedding_directory=None, clip
             if clip_type == CLIPType.HUNYUAN_IMAGE:
                 clip_target.clip = comfy.text_encoders.hunyuan_image.te(byt5=False, **llama_detect(clip_data))
                 clip_target.tokenizer = comfy.text_encoders.hunyuan_image.HunyuanImageTokenizer
+            elif clip_type == CLIPType.LONGCAT_IMAGE:
+                clip_target.clip = comfy.text_encoders.longcat_image.te(**llama_detect(clip_data))
+                clip_target.tokenizer = comfy.text_encoders.longcat_image.LongCatImageTokenizer
             else:
                 clip_target.clip = comfy.text_encoders.qwen_image.te(**llama_detect(clip_data))
                 clip_target.tokenizer = comfy.text_encoders.qwen_image.QwenImageTokenizer

comfy/supported_models.py

@@ -25,6 +25,7 @@ import comfy.text_encoders.kandinsky5
 import comfy.text_encoders.z_image
 import comfy.text_encoders.anima
 import comfy.text_encoders.ace15
+import comfy.text_encoders.longcat_image
 
 from . import supported_models_base
 from . import latent_formats
@@ -525,7 +526,8 @@ class LotusD(SD20):
     }
 
     unet_extra_config = {
-        "num_classes": 'sequential'
+        "num_classes": 'sequential',
+        "num_head_channels": 64,
     }
 
     def get_model(self, state_dict, prefix="", device=None):
@@ -1677,6 +1679,37 @@ class ACEStep15(supported_models_base.BASE):
         return supported_models_base.ClipTarget(comfy.text_encoders.ace15.ACE15Tokenizer, comfy.text_encoders.ace15.te(**detect))
 
 
-models = [LotusD, Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p, SD3, StableAudio, AuraFlow, PixArtAlpha, PixArtSigma, HunyuanDiT, HunyuanDiT1, FluxInpaint, Flux, FluxSchnell, GenmoMochi, LTXV, LTXAV, HunyuanVideo15_SR_Distilled, HunyuanVideo15, HunyuanImage21Refiner, HunyuanImage21, HunyuanVideoSkyreelsI2V, HunyuanVideoI2V, HunyuanVideo, CosmosT2V, CosmosI2V, CosmosT2IPredict2, CosmosI2VPredict2, ZImage, Lumina2, WAN22_T2V, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, WAN21_Vace, WAN21_Camera, WAN22_Camera, WAN22_S2V, WAN21_HuMo, WAN22_Animate, WAN21_FlowRVS, Hunyuan3Dv2mini, Hunyuan3Dv2, Hunyuan3Dv2_1, HiDream, Chroma, ChromaRadiance, ACEStep, ACEStep15, Omnigen2, QwenImage, Flux2, Kandinsky5Image, Kandinsky5, Anima]
+class LongCatImage(supported_models_base.BASE):
+    unet_config = {
+        "image_model": "flux",
+        "guidance_embed": False,
+        "vec_in_dim": None,
+        "context_in_dim": 3584,
+        "txt_ids_dims": [1, 2],
+    }
+
+    sampling_settings = {
+    }
+
+    unet_extra_config = {}
+    latent_format = latent_formats.Flux
+
+    memory_usage_factor = 2.5
+
+    supported_inference_dtypes = [torch.bfloat16, torch.float16, torch.float32]
+
+    vae_key_prefix = ["vae."]
+    text_encoder_key_prefix = ["text_encoders."]
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.LongCatImage(self, device=device)
+        return out
+
+    def clip_target(self, state_dict={}):
+        pref = self.text_encoder_key_prefix[0]
+        hunyuan_detect = comfy.text_encoders.hunyuan_video.llama_detect(state_dict, "{}qwen25_7b.transformer.".format(pref))
+        return supported_models_base.ClipTarget(comfy.text_encoders.longcat_image.LongCatImageTokenizer, comfy.text_encoders.longcat_image.te(**hunyuan_detect))
+
+models = [LotusD, Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p, SD3, StableAudio, AuraFlow, PixArtAlpha, PixArtSigma, HunyuanDiT, HunyuanDiT1, FluxInpaint, Flux, LongCatImage, FluxSchnell, GenmoMochi, LTXV, LTXAV, HunyuanVideo15_SR_Distilled, HunyuanVideo15, HunyuanImage21Refiner, HunyuanImage21, HunyuanVideoSkyreelsI2V, HunyuanVideoI2V, HunyuanVideo, CosmosT2V, CosmosI2V, CosmosT2IPredict2, CosmosI2VPredict2, ZImage, Lumina2, WAN22_T2V, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, WAN21_Vace, WAN21_Camera, WAN22_Camera, WAN22_S2V, WAN21_HuMo, WAN22_Animate, WAN21_FlowRVS, Hunyuan3Dv2mini, Hunyuan3Dv2, Hunyuan3Dv2_1, HiDream, Chroma, ChromaRadiance, ACEStep, ACEStep15, Omnigen2, QwenImage, Flux2, Kandinsky5Image, Kandinsky5, Anima]
 
 models += [SVD_img2vid]

comfy/text_encoders/ace15.py

@@ -328,14 +328,14 @@ class ACE15TEModel(torch.nn.Module):
                 return getattr(self, self.lm_model).load_sd(sd)
 
     def memory_estimation_function(self, token_weight_pairs, device=None):
-        lm_metadata = token_weight_pairs["lm_metadata"]
+        lm_metadata = token_weight_pairs.get("lm_metadata", {})
         constant = self.constant
         if comfy.model_management.should_use_bf16(device):
             constant *= 0.5
 
         token_weight_pairs = token_weight_pairs.get("lm_prompt", [])
         num_tokens = sum(map(lambda a: len(a), token_weight_pairs))
-        num_tokens += lm_metadata['min_tokens']
+        num_tokens += lm_metadata.get("min_tokens", 0)
         return num_tokens * constant * 1024 * 1024
 
 def te(dtype_llama=None, llama_quantization_metadata=None, lm_model="qwen3_2b"):

comfy/text_encoders/longcat_image.py

@@ -0,0 +1,184 @@
+import re
+import numbers
+import torch
+from comfy import sd1_clip
+from comfy.text_encoders.qwen_image import Qwen25_7BVLITokenizer, Qwen25_7BVLIModel
+import logging
+
+logger = logging.getLogger(__name__)
+
+QUOTE_PAIRS = [("'", "'"), ('"', '"'), ("\u2018", "\u2019"), ("\u201c", "\u201d")]
+QUOTE_PATTERN = "|".join(
+    [
+        re.escape(q1) + r"[^" + re.escape(q1 + q2) + r"]*?" + re.escape(q2)
+        for q1, q2 in QUOTE_PAIRS
+    ]
+)
+WORD_INTERNAL_QUOTE_RE = re.compile(r"[a-zA-Z]+'[a-zA-Z]+")
+
+
+def split_quotation(prompt):
+    matches = WORD_INTERNAL_QUOTE_RE.findall(prompt)
+    mapping = []
+    for i, word_src in enumerate(set(matches)):
+        word_tgt = "longcat_$##$_longcat" * (i + 1)
+        prompt = prompt.replace(word_src, word_tgt)
+        mapping.append((word_src, word_tgt))
+
+    parts = re.split(f"({QUOTE_PATTERN})", prompt)
+    result = []
+    for part in parts:
+        for word_src, word_tgt in mapping:
+            part = part.replace(word_tgt, word_src)
+        if not part:
+            continue
+        is_quoted = bool(re.match(QUOTE_PATTERN, part))
+        result.append((part, is_quoted))
+    return result
+
+
+class LongCatImageBaseTokenizer(Qwen25_7BVLITokenizer):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.max_length = 512
+
+    def tokenize_with_weights(self, text, return_word_ids=False, **kwargs):
+        parts = split_quotation(text)
+        all_tokens = []
+        for part_text, is_quoted in parts:
+            if is_quoted:
+                for char in part_text:
+                    ids = self.tokenizer(char, add_special_tokens=False)["input_ids"]
+                    all_tokens.extend(ids)
+            else:
+                ids = self.tokenizer(part_text, add_special_tokens=False)["input_ids"]
+                all_tokens.extend(ids)
+
+        if len(all_tokens) > self.max_length:
+            all_tokens = all_tokens[: self.max_length]
+            logger.warning(f"Truncated prompt to {self.max_length} tokens")
+
+        output = [(t, 1.0) for t in all_tokens]
+        # Pad to max length
+        self.pad_tokens(output, self.max_length - len(output))
+        return [output]
+
+
+class LongCatImageTokenizer(sd1_clip.SD1Tokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        super().__init__(
+            embedding_directory=embedding_directory,
+            tokenizer_data=tokenizer_data,
+            name="qwen25_7b",
+            tokenizer=LongCatImageBaseTokenizer,
+        )
+        self.longcat_template_prefix = "<|im_start|>system\nAs an image captioning expert, generate a descriptive text prompt based on an image content, suitable for input to a text-to-image model.<|im_end|>\n<|im_start|>user\n"
+        self.longcat_template_suffix = "<|im_end|>\n<|im_start|>assistant\n"
+
+    def tokenize_with_weights(self, text, return_word_ids=False, **kwargs):
+        skip_template = False
+        if text.startswith("<|im_start|>"):
+            skip_template = True
+        if text.startswith("<|start_header_id|>"):
+            skip_template = True
+        if text == "":
+            text = " "
+
+        base_tok = getattr(self, "qwen25_7b")
+        if skip_template:
+            tokens = super().tokenize_with_weights(
+                text, return_word_ids=return_word_ids, disable_weights=True, **kwargs
+            )
+        else:
+            prefix_ids = base_tok.tokenizer(
+                self.longcat_template_prefix, add_special_tokens=False
+            )["input_ids"]
+            suffix_ids = base_tok.tokenizer(
+                self.longcat_template_suffix, add_special_tokens=False
+            )["input_ids"]
+
+            prompt_tokens = base_tok.tokenize_with_weights(
+                text, return_word_ids=return_word_ids, **kwargs
+            )
+            prompt_pairs = prompt_tokens[0]
+
+            prefix_pairs = [(t, 1.0) for t in prefix_ids]
+            suffix_pairs = [(t, 1.0) for t in suffix_ids]
+
+            combined = prefix_pairs + prompt_pairs + suffix_pairs
+            tokens = {"qwen25_7b": [combined]}
+
+        return tokens
+
+
+class LongCatImageTEModel(sd1_clip.SD1ClipModel):
+    def __init__(self, device="cpu", dtype=None, model_options={}):
+        super().__init__(
+            device=device,
+            dtype=dtype,
+            name="qwen25_7b",
+            clip_model=Qwen25_7BVLIModel,
+            model_options=model_options,
+        )
+
+    def encode_token_weights(self, token_weight_pairs, template_end=-1):
+        out, pooled, extra = super().encode_token_weights(token_weight_pairs)
+        tok_pairs = token_weight_pairs["qwen25_7b"][0]
+        count_im_start = 0
+        if template_end == -1:
+            for i, v in enumerate(tok_pairs):
+                elem = v[0]
+                if not torch.is_tensor(elem):
+                    if isinstance(elem, numbers.Integral):
+                        if elem == 151644 and count_im_start < 2:
+                            template_end = i
+                            count_im_start += 1
+
+        if out.shape[1] > (template_end + 3):
+            if tok_pairs[template_end + 1][0] == 872:
+                if tok_pairs[template_end + 2][0] == 198:
+                    template_end += 3
+
+        if template_end == -1:
+            template_end = 0
+
+        suffix_start = None
+        for i in range(len(tok_pairs) - 1, -1, -1):
+            elem = tok_pairs[i][0]
+            if not torch.is_tensor(elem) and isinstance(elem, numbers.Integral):
+                if elem == 151645:
+                    suffix_start = i
+                    break
+
+        out = out[:, template_end:]
+
+        if "attention_mask" in extra:
+            extra["attention_mask"] = extra["attention_mask"][:, template_end:]
+            if extra["attention_mask"].sum() == torch.numel(extra["attention_mask"]):
+                extra.pop("attention_mask")
+
+        if suffix_start is not None:
+            suffix_len = len(tok_pairs) - suffix_start
+            if suffix_len > 0 and out.shape[1] > suffix_len:
+                out = out[:, :-suffix_len]
+                if "attention_mask" in extra:
+                    extra["attention_mask"] = extra["attention_mask"][:, :-suffix_len]
+                    if extra["attention_mask"].sum() == torch.numel(
+                        extra["attention_mask"]
+                    ):
+                        extra.pop("attention_mask")
+
+        return out, pooled, extra
+
+
+def te(dtype_llama=None, llama_quantization_metadata=None):
+    class LongCatImageTEModel_(LongCatImageTEModel):
+        def __init__(self, device="cpu", dtype=None, model_options={}):
+            if llama_quantization_metadata is not None:
+                model_options = model_options.copy()
+                model_options["quantization_metadata"] = llama_quantization_metadata
+            if dtype_llama is not None:
+                dtype = dtype_llama
+            super().__init__(device=device, dtype=dtype, model_options=model_options)
+
+    return LongCatImageTEModel_

comfy_api/latest/_io.py

@@ -1224,9 +1224,10 @@ class BoundingBox(ComfyTypeIO):
 
     class Input(WidgetInput):
         def __init__(self, id: str, display_name: str=None, optional=False, tooltip: str=None,
-                     socketless: bool=True, default: dict=None, component: str=None):
+                     socketless: bool=True, default: dict=None, component: str=None, force_input: bool=None):
             super().__init__(id, display_name, optional, tooltip, None, default, socketless)
             self.component = component
+            self.force_input = force_input
             if default is None:
                 self.default = {"x": 0, "y": 0, "width": 512, "height": 512}
 
@@ -1234,6 +1235,8 @@ class BoundingBox(ComfyTypeIO):
             d = super().as_dict()
             if self.component:
                 d["component"] = self.component
+            if self.force_input is not None:
+                d["forceInput"] = self.force_input
             return d
 
 

comfy_extras/nodes_audio.py

@@ -162,7 +162,7 @@ class SaveAudio(IO.ComfyNode):
             essentials_category="Audio",
             inputs=[
                 IO.Audio.Input("audio"),
-                IO.String.Input("filename_prefix", default="audio/ComfyUI"),
+                IO.String.Input("filename_prefix", default="audio/ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
             ],
             hidden=[IO.Hidden.prompt, IO.Hidden.extra_pnginfo],
             is_output_node=True,
@@ -187,7 +187,7 @@ class SaveAudioMP3(IO.ComfyNode):
             category="audio",
             inputs=[
                 IO.Audio.Input("audio"),
-                IO.String.Input("filename_prefix", default="audio/ComfyUI"),
+                IO.String.Input("filename_prefix", default="audio/ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
                 IO.Combo.Input("quality", options=["V0", "128k", "320k"], default="V0"),
             ],
             hidden=[IO.Hidden.prompt, IO.Hidden.extra_pnginfo],
@@ -215,7 +215,7 @@ class SaveAudioOpus(IO.ComfyNode):
             category="audio",
             inputs=[
                 IO.Audio.Input("audio"),
-                IO.String.Input("filename_prefix", default="audio/ComfyUI"),
+                IO.String.Input("filename_prefix", default="audio/ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
                 IO.Combo.Input("quality", options=["64k", "96k", "128k", "192k", "320k"], default="128k"),
             ],
             hidden=[IO.Hidden.prompt, IO.Hidden.extra_pnginfo],

comfy_extras/nodes_glsl.py

@@ -865,14 +865,15 @@ class GLSLShader(io.ComfyNode):
         cls, image_list: list[torch.Tensor], output_batch: torch.Tensor
     ) -> dict[str, list]:
         """Build UI output with input and output images for client-side shader execution."""
-        combined_inputs = torch.cat(image_list, dim=0)
-        input_images_ui = ui.ImageSaveHelper.save_images(
-            combined_inputs,
-            filename_prefix="GLSLShader_input",
-            folder_type=io.FolderType.temp,
-            cls=None,
-            compress_level=1,
-        )
+        input_images_ui = []
+        for img in image_list:
+            input_images_ui.extend(ui.ImageSaveHelper.save_images(
+                img,
+                filename_prefix="GLSLShader_input",
+                folder_type=io.FolderType.temp,
+                cls=None,
+                compress_level=1,
+            ))
 
         output_images_ui = ui.ImageSaveHelper.save_images(
             output_batch,

comfy_extras/nodes_hunyuan3d.py

@@ -637,7 +637,7 @@ class SaveGLB(IO.ComfyNode):
                     ],
                     tooltip="Mesh or 3D file to save",
                 ),
-                IO.String.Input("filename_prefix", default="mesh/ComfyUI"),
+                IO.String.Input("filename_prefix", default="mesh/ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
             ],
             hidden=[IO.Hidden.prompt, IO.Hidden.extra_pnginfo]
         )

comfy_extras/nodes_images.py

@@ -190,7 +190,7 @@ class SaveAnimatedWEBP(IO.ComfyNode):
             category="image/animation",
             inputs=[
                 IO.Image.Input("images"),
-                IO.String.Input("filename_prefix", default="ComfyUI"),
+                IO.String.Input("filename_prefix", default="ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
                 IO.Float.Input("fps", default=6.0, min=0.01, max=1000.0, step=0.01),
                 IO.Boolean.Input("lossless", default=True),
                 IO.Int.Input("quality", default=80, min=0, max=100),
@@ -227,7 +227,7 @@ class SaveAnimatedPNG(IO.ComfyNode):
             category="image/animation",
             inputs=[
                 IO.Image.Input("images"),
-                IO.String.Input("filename_prefix", default="ComfyUI"),
+                IO.String.Input("filename_prefix", default="ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
                 IO.Float.Input("fps", default=6.0, min=0.01, max=1000.0, step=0.01),
                 IO.Int.Input("compress_level", default=4, min=0, max=9, advanced=True),
             ],
@@ -489,7 +489,7 @@ class SaveSVGNode(IO.ComfyNode):
                 IO.SVG.Input("svg"),
                 IO.String.Input(
                     "filename_prefix",
-                    default="svg/ComfyUI",
+                    default="svg/ComfyUI_%year%%month%%day%-%hour%%minute%%second%",
                     tooltip="The prefix for the file to save. This may include formatting information such as %date:yyyy-MM-dd% or %Empty Latent Image.width% to include values from nodes.",
                 ),
             ],
@@ -706,8 +706,8 @@ class SplitImageToTileList(IO.ComfyNode):
     @staticmethod
     def get_grid_coords(width, height, tile_width, tile_height, overlap):
         coords = []
-        stride_x = max(1, tile_width - overlap)
-        stride_y = max(1, tile_height - overlap)
+        stride_x = round(max(tile_width * 0.25, tile_width - overlap))
+        stride_y = round(max(tile_width * 0.25, tile_height - overlap))
 
         y = 0
         while y < height:
@@ -764,34 +764,6 @@ class ImageMergeTileList(IO.ComfyNode):
             ],
         )
 
-    @staticmethod
-    def get_grid_coords(width, height, tile_width, tile_height, overlap):
-        coords = []
-        stride_x = max(1, tile_width - overlap)
-        stride_y = max(1, tile_height - overlap)
-
-        y = 0
-        while y < height:
-            x = 0
-            y_end = min(y + tile_height, height)
-            y_start = max(0, y_end - tile_height)
-
-            while x < width:
-                x_end = min(x + tile_width, width)
-                x_start = max(0, x_end - tile_width)
-
-                coords.append((x_start, y_start, x_end, y_end))
-
-                if x_end >= width:
-                    break
-                x += stride_x
-
-            if y_end >= height:
-                break
-            y += stride_y
-
-        return coords
-
     @classmethod
     def execute(cls, image_list, final_width, final_height, overlap):
         w = final_width[0]
@@ -804,7 +776,7 @@ class ImageMergeTileList(IO.ComfyNode):
         device = first_tile.device
         dtype = first_tile.dtype
 
-        coords = cls.get_grid_coords(w, h, t_w, t_h, ovlp)
+        coords = SplitImageToTileList.get_grid_coords(w, h, t_w, t_h, ovlp)
 
         canvas = torch.zeros((b, h, w, c), device=device, dtype=dtype)
         weights = torch.zeros((b, h, w, 1), device=device, dtype=dtype)

comfy_extras/nodes_resolution.py

@@ -0,0 +1,86 @@
+from __future__ import annotations
+import math
+from enum import Enum
+from typing_extensions import override
+from comfy_api.latest import ComfyExtension, io
+
+
+class AspectRatio(str, Enum):
+    SQUARE = "1:1 (Square)"
+    PHOTO_H = "3:2 (Photo)"
+    STANDARD_H = "4:3 (Standard)"
+    WIDESCREEN_H = "16:9 (Widescreen)"
+    ULTRAWIDE_H = "21:9 (Ultrawide)"
+    PHOTO_V = "2:3 (Portrait Photo)"
+    STANDARD_V = "3:4 (Portrait Standard)"
+    WIDESCREEN_V = "9:16 (Portrait Widescreen)"
+
+
+ASPECT_RATIOS: dict[AspectRatio, tuple[int, int]] = {
+    AspectRatio.SQUARE: (1, 1),
+    AspectRatio.PHOTO_H: (3, 2),
+    AspectRatio.STANDARD_H: (4, 3),
+    AspectRatio.WIDESCREEN_H: (16, 9),
+    AspectRatio.ULTRAWIDE_H: (21, 9),
+    AspectRatio.PHOTO_V: (2, 3),
+    AspectRatio.STANDARD_V: (3, 4),
+    AspectRatio.WIDESCREEN_V: (9, 16),
+}
+
+
+class ResolutionSelector(io.ComfyNode):
+    """Calculate width and height from aspect ratio and megapixel target."""
+
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ResolutionSelector",
+            display_name="Resolution Selector",
+            category="utils",
+            description="Calculate width and height from aspect ratio and megapixel target. Useful for setting up Empty Latent Image dimensions.",
+            inputs=[
+                io.Combo.Input(
+                    "aspect_ratio",
+                    options=AspectRatio,
+                    default=AspectRatio.SQUARE,
+                    tooltip="The aspect ratio for the output dimensions.",
+                ),
+                io.Float.Input(
+                    "megapixels",
+                    default=1.0,
+                    min=0.1,
+                    max=16.0,
+                    step=0.1,
+                    tooltip="Target total megapixels. 1.0 MP ≈ 1024×1024 for square.",
+                ),
+            ],
+            outputs=[
+                io.Int.Output(
+                    "width", tooltip="Calculated width in pixels (multiple of 8)."
+                ),
+                io.Int.Output(
+                    "height", tooltip="Calculated height in pixels (multiple of 8)."
+                ),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, aspect_ratio: str, megapixels: float) -> io.NodeOutput:
+        w_ratio, h_ratio = ASPECT_RATIOS[aspect_ratio]
+        total_pixels = megapixels * 1024 * 1024
+        scale = math.sqrt(total_pixels / (w_ratio * h_ratio))
+        width = round(w_ratio * scale / 8) * 8
+        height = round(h_ratio * scale / 8) * 8
+        return io.NodeOutput(width, height)
+
+
+class ResolutionExtension(ComfyExtension):
+    @override
+    async def get_node_list(self) -> list[type[io.ComfyNode]]:
+        return [
+            ResolutionSelector,
+        ]
+
+
+async def comfy_entrypoint() -> ResolutionExtension:
+    return ResolutionExtension()

comfy_extras/nodes_sdpose.py

@@ -0,0 +1,740 @@
+import torch
+import comfy.utils
+import numpy as np
+import math
+import colorsys
+from tqdm import tqdm
+from typing_extensions import override
+from comfy_api.latest import ComfyExtension, io
+from comfy_extras.nodes_lotus import LotusConditioning
+
+
+def _preprocess_keypoints(kp_raw, sc_raw):
+    """Insert neck keypoint and remap from MMPose to OpenPose ordering.
+
+    Returns (kp, sc) where kp has shape (134, 2) and sc has shape (134,).
+    Layout:
+      0-17   body  (18 kp, OpenPose order)
+      18-23  feet  (6 kp)
+      24-91  face  (68 kp)
+      92-112 right hand (21 kp)
+      113-133 left hand (21 kp)
+    """
+    kp = np.array(kp_raw, dtype=np.float32)
+    sc = np.array(sc_raw, dtype=np.float32)
+    if len(kp) >= 17:
+        neck = (kp[5] + kp[6]) / 2
+        neck_score = min(sc[5], sc[6]) if sc[5] > 0.3 and sc[6] > 0.3 else 0
+        kp = np.insert(kp, 17, neck, axis=0)
+        sc = np.insert(sc, 17, neck_score)
+        mmpose_idx   = np.array([17, 6,  8, 10,  7,  9, 12, 14, 16, 13, 15, 2, 1, 4, 3])
+        openpose_idx = np.array([ 1, 2,  3,  4,  6,  7,  8,  9, 10, 12, 13, 14, 15, 16, 17])
+        tmp_kp, tmp_sc = kp.copy(), sc.copy()
+        tmp_kp[openpose_idx] = kp[mmpose_idx]
+        tmp_sc[openpose_idx] = sc[mmpose_idx]
+        kp, sc = tmp_kp, tmp_sc
+    return kp, sc
+
+
+def _to_openpose_frames(all_keypoints, all_scores, height, width):
+    """Convert raw keypoint lists to a list of OpenPose-style frame dicts.
+
+    Each frame dict contains:
+      canvas_width, canvas_height, people: list of person dicts with keys:
+        pose_keypoints_2d       - 18 body kp  as flat [x,y,score,...] (absolute pixels)
+        foot_keypoints_2d       -  6 foot kp  as flat [x,y,score,...] (absolute pixels)
+        face_keypoints_2d       - 70 face kp  as flat [x,y,score,...] (absolute pixels)
+                                   indices 0-67: 68 face landmarks
+                                   index  68:    right eye (body[14])
+                                   index  69:    left  eye (body[15])
+        hand_right_keypoints_2d - 21 right-hand kp (absolute pixels)
+        hand_left_keypoints_2d  - 21 left-hand  kp (absolute pixels)
+    """
+    def _flatten(kp_slice, sc_slice):
+        return np.stack([kp_slice[:, 0], kp_slice[:, 1], sc_slice], axis=1).flatten().tolist()
+
+    frames = []
+    for img_idx in range(len(all_keypoints)):
+        people = []
+        for kp_raw, sc_raw in zip(all_keypoints[img_idx], all_scores[img_idx]):
+            kp, sc = _preprocess_keypoints(kp_raw, sc_raw)
+            # 70 face kp = 68 face landmarks + REye (body[14]) + LEye (body[15])
+            face_kp = np.concatenate([kp[24:92], kp[[14, 15]]], axis=0)
+            face_sc = np.concatenate([sc[24:92], sc[[14, 15]]], axis=0)
+            people.append({
+                "pose_keypoints_2d":       _flatten(kp[0:18],   sc[0:18]),
+                "foot_keypoints_2d":       _flatten(kp[18:24],  sc[18:24]),
+                "face_keypoints_2d":       _flatten(face_kp,    face_sc),
+                "hand_right_keypoints_2d": _flatten(kp[92:113], sc[92:113]),
+                "hand_left_keypoints_2d":  _flatten(kp[113:134], sc[113:134]),
+            })
+        frames.append({"canvas_width": width, "canvas_height": height, "people": people})
+    return frames
+
+
+class KeypointDraw:
+    """
+    Pose keypoint drawing class that supports both numpy and cv2 backends.
+    """
+    def __init__(self):
+        try:
+            import cv2
+            self.draw = cv2
+        except ImportError:
+            self.draw = self
+
+        # Hand connections (same for both hands)
+        self.hand_edges = [
+            [0, 1], [1, 2], [2, 3], [3, 4],      # thumb
+            [0, 5], [5, 6], [6, 7], [7, 8],      # index
+            [0, 9], [9, 10], [10, 11], [11, 12], # middle
+            [0, 13], [13, 14], [14, 15], [15, 16], # ring
+            [0, 17], [17, 18], [18, 19], [19, 20], # pinky
+        ]
+
+        # Body connections - matching DWPose limbSeq (1-indexed, converted to 0-indexed)
+        self.body_limbSeq = [
+            [2, 3], [2, 6], [3, 4], [4, 5], [6, 7], [7, 8], [2, 9], [9, 10],
+            [10, 11], [2, 12], [12, 13], [13, 14], [2, 1], [1, 15], [15, 17],
+            [1, 16], [16, 18]
+        ]
+
+        # Colors matching DWPose
+        self.colors = [
+            [255, 0, 0], [255, 85, 0], [255, 170, 0], [255, 255, 0], [170, 255, 0],
+            [85, 255, 0], [0, 255, 0], [0, 255, 85], [0, 255, 170], [0, 255, 255],
+            [0, 170, 255], [0, 85, 255], [0, 0, 255], [85, 0, 255],
+            [170, 0, 255], [255, 0, 255], [255, 0, 170], [255, 0, 85]
+        ]
+
+    @staticmethod
+    def circle(canvas_np, center, radius, color, **kwargs):
+        """Draw a filled circle using NumPy vectorized operations."""
+        cx, cy = center
+        h, w = canvas_np.shape[:2]
+
+        radius_int = int(np.ceil(radius))
+
+        y_min, y_max = max(0, cy - radius_int), min(h, cy + radius_int + 1)
+        x_min, x_max = max(0, cx - radius_int), min(w, cx + radius_int + 1)
+
+        if y_max <= y_min or x_max <= x_min:
+            return
+
+        y, x = np.ogrid[y_min:y_max, x_min:x_max]
+        mask = (x - cx)**2 + (y - cy)**2 <= radius**2
+        canvas_np[y_min:y_max, x_min:x_max][mask] = color
+
+    @staticmethod
+    def line(canvas_np, pt1, pt2, color, thickness=1, **kwargs):
+        """Draw line using Bresenham's algorithm with NumPy operations."""
+        x0, y0, x1, y1 = *pt1, *pt2
+        h, w = canvas_np.shape[:2]
+        dx, dy = abs(x1 - x0), abs(y1 - y0)
+        sx, sy = (1 if x0 < x1 else -1), (1 if y0 < y1 else -1)
+        err, x, y, line_points = dx - dy, x0, y0, []
+
+        while True:
+            line_points.append((x, y))
+            if x == x1 and y == y1:
+                break
+            e2 = 2 * err
+            if e2 > -dy:
+                err, x = err - dy, x + sx
+            if e2 < dx:
+                err, y = err + dx, y + sy
+
+        if thickness > 1:
+            radius, radius_int = (thickness / 2.0) + 0.5, int(np.ceil((thickness / 2.0) + 0.5))
+            for px, py in line_points:
+                y_min, y_max, x_min, x_max = max(0, py - radius_int), min(h, py + radius_int + 1), max(0, px - radius_int), min(w, px + radius_int + 1)
+                if y_max > y_min and x_max > x_min:
+                    yy, xx = np.ogrid[y_min:y_max, x_min:x_max]
+                    canvas_np[y_min:y_max, x_min:x_max][(xx - px)**2 + (yy - py)**2 <= radius**2] = color
+        else:
+            line_points = np.array(line_points)
+            valid = (line_points[:, 1] >= 0) & (line_points[:, 1] < h) & (line_points[:, 0] >= 0) & (line_points[:, 0] < w)
+            if (valid_points := line_points[valid]).size:
+                canvas_np[valid_points[:, 1], valid_points[:, 0]] = color
+
+    @staticmethod
+    def fillConvexPoly(canvas_np, pts, color, **kwargs):
+        """Fill polygon using vectorized scanline algorithm."""
+        if len(pts) < 3:
+            return
+        pts = np.array(pts, dtype=np.int32)
+        h, w = canvas_np.shape[:2]
+        y_min, y_max, x_min, x_max = max(0, pts[:, 1].min()), min(h, pts[:, 1].max() + 1), max(0, pts[:, 0].min()), min(w, pts[:, 0].max() + 1)
+        if y_max <= y_min or x_max <= x_min:
+            return
+        yy, xx = np.mgrid[y_min:y_max, x_min:x_max]
+        mask = np.zeros((y_max - y_min, x_max - x_min), dtype=bool)
+
+        for i in range(len(pts)):
+            p1, p2 = pts[i], pts[(i + 1) % len(pts)]
+            y1, y2 = p1[1], p2[1]
+            if y1 == y2:
+                continue
+            if y1 > y2:
+                p1, p2, y1, y2 = p2, p1, p2[1], p1[1]
+            if not (edge_mask := (yy >= y1) & (yy < y2)).any():
+                continue
+            mask ^= edge_mask & (xx >= p1[0] + (yy - y1) * (p2[0] - p1[0]) / (y2 - y1))
+
+        canvas_np[y_min:y_max, x_min:x_max][mask] = color
+
+    @staticmethod
+    def ellipse2Poly(center, axes, angle, arc_start, arc_end, delta=1, **kwargs):
+        """Python implementation of cv2.ellipse2Poly."""
+        axes = (axes[0] + 0.5, axes[1] + 0.5) # to better match cv2 output
+        angle = angle % 360
+        if arc_start > arc_end:
+            arc_start, arc_end = arc_end, arc_start
+        while arc_start < 0:
+            arc_start, arc_end = arc_start + 360, arc_end + 360
+        while arc_end > 360:
+            arc_end, arc_start = arc_end - 360, arc_start - 360
+        if arc_end - arc_start > 360:
+            arc_start, arc_end = 0, 360
+
+        angle_rad = math.radians(angle)
+        alpha, beta = math.cos(angle_rad), math.sin(angle_rad)
+        pts = []
+        for i in range(arc_start, arc_end + delta, delta):
+            theta_rad = math.radians(min(i, arc_end))
+            x, y = axes[0] * math.cos(theta_rad), axes[1] * math.sin(theta_rad)
+            pts.append([int(round(center[0] + x * alpha - y * beta)), int(round(center[1] + x * beta + y * alpha))])
+
+        unique_pts, prev_pt = [], (float('inf'), float('inf'))
+        for pt in pts:
+            if (pt_tuple := tuple(pt)) != prev_pt:
+                unique_pts.append(pt)
+                prev_pt = pt_tuple
+
+        return unique_pts if len(unique_pts) > 1 else [[center[0], center[1]], [center[0], center[1]]]
+
+    def draw_wholebody_keypoints(self, canvas, keypoints, scores=None, threshold=0.3,
+                                 draw_body=True, draw_feet=True, draw_face=True, draw_hands=True, stick_width=4, face_point_size=3):
+        """
+        Draw wholebody keypoints (134 keypoints after processing) in DWPose style.
+
+        Expected keypoint format (after neck insertion and remapping):
+        - Body: 0-17 (18 keypoints in OpenPose format, neck at index 1)
+        - Foot: 18-23 (6 keypoints)
+        - Face: 24-91 (68 landmarks)
+        - Right hand: 92-112 (21 keypoints)
+        - Left hand: 113-133 (21 keypoints)
+
+        Args:
+            canvas: The canvas to draw on (numpy array)
+            keypoints: Array of keypoint coordinates
+            scores: Optional confidence scores for each keypoint
+            threshold: Minimum confidence threshold for drawing keypoints
+
+        Returns:
+            canvas: The canvas with keypoints drawn
+        """
+        H, W, C = canvas.shape
+
+        # Draw body limbs
+        if draw_body and len(keypoints) >= 18:
+            for i, limb in enumerate(self.body_limbSeq):
+                # Convert from 1-indexed to 0-indexed
+                idx1, idx2 = limb[0] - 1, limb[1] - 1
+
+                if idx1 >= 18 or idx2 >= 18:
+                    continue
+
+                if scores is not None:
+                    if scores[idx1] < threshold or scores[idx2] < threshold:
+                        continue
+
+                Y = [keypoints[idx1][0], keypoints[idx2][0]]
+                X = [keypoints[idx1][1], keypoints[idx2][1]]
+                mX, mY = (X[0] + X[1]) / 2, (Y[0] + Y[1]) / 2
+                length = math.sqrt((X[0] - X[1]) ** 2 + (Y[0] - Y[1]) ** 2)
+
+                if length < 1:
+                    continue
+
+                angle = math.degrees(math.atan2(X[0] - X[1], Y[0] - Y[1]))
+
+                polygon = self.draw.ellipse2Poly((int(mY), int(mX)), (int(length / 2), stick_width), int(angle), 0, 360, 1)
+
+                self.draw.fillConvexPoly(canvas, polygon, self.colors[i % len(self.colors)])
+
+        # Draw body keypoints
+        if draw_body and len(keypoints) >= 18:
+            for i in range(18):
+                if scores is not None and scores[i] < threshold:
+                    continue
+                x, y = int(keypoints[i][0]), int(keypoints[i][1])
+                if 0 <= x < W and 0 <= y < H:
+                    self.draw.circle(canvas, (x, y), 4, self.colors[i % len(self.colors)], thickness=-1)
+
+        # Draw foot keypoints (18-23, 6 keypoints)
+        if draw_feet and len(keypoints) >= 24:
+            for i in range(18, 24):
+                if scores is not None and scores[i] < threshold:
+                    continue
+                x, y = int(keypoints[i][0]), int(keypoints[i][1])
+                if 0 <= x < W and 0 <= y < H:
+                    self.draw.circle(canvas, (x, y), 4, self.colors[i % len(self.colors)], thickness=-1)
+
+        # Draw right hand (92-112)
+        if draw_hands and len(keypoints) >= 113:
+            eps = 0.01
+            for ie, edge in enumerate(self.hand_edges):
+                idx1, idx2 = 92 + edge[0], 92 + edge[1]
+                if scores is not None:
+                    if scores[idx1] < threshold or scores[idx2] < threshold:
+                        continue
+
+                x1, y1 = int(keypoints[idx1][0]), int(keypoints[idx1][1])
+                x2, y2 = int(keypoints[idx2][0]), int(keypoints[idx2][1])
+
+                if x1 > eps and y1 > eps and x2 > eps and y2 > eps:
+                    if 0 <= x1 < W and 0 <= y1 < H and 0 <= x2 < W and 0 <= y2 < H:
+                        # HSV to RGB conversion for rainbow colors
+                        r, g, b = colorsys.hsv_to_rgb(ie / float(len(self.hand_edges)), 1.0, 1.0)
+                        color = (int(r * 255), int(g * 255), int(b * 255))
+                        self.draw.line(canvas, (x1, y1), (x2, y2), color, thickness=2)
+
+            # Draw right hand keypoints
+            for i in range(92, 113):
+                if scores is not None and scores[i] < threshold:
+                    continue
+                x, y = int(keypoints[i][0]), int(keypoints[i][1])
+                if x > eps and y > eps and 0 <= x < W and 0 <= y < H:
+                    self.draw.circle(canvas, (x, y), 4, (0, 0, 255), thickness=-1)
+
+        # Draw left hand (113-133)
+        if draw_hands and len(keypoints) >= 134:
+            eps = 0.01
+            for ie, edge in enumerate(self.hand_edges):
+                idx1, idx2 = 113 + edge[0], 113 + edge[1]
+                if scores is not None:
+                    if scores[idx1] < threshold or scores[idx2] < threshold:
+                        continue
+
+                x1, y1 = int(keypoints[idx1][0]), int(keypoints[idx1][1])
+                x2, y2 = int(keypoints[idx2][0]), int(keypoints[idx2][1])
+
+                if x1 > eps and y1 > eps and x2 > eps and y2 > eps:
+                    if 0 <= x1 < W and 0 <= y1 < H and 0 <= x2 < W and 0 <= y2 < H:
+                        # HSV to RGB conversion for rainbow colors
+                        r, g, b = colorsys.hsv_to_rgb(ie / float(len(self.hand_edges)), 1.0, 1.0)
+                        color = (int(r * 255), int(g * 255), int(b * 255))
+                        self.draw.line(canvas, (x1, y1), (x2, y2), color, thickness=2)
+
+            # Draw left hand keypoints
+            for i in range(113, 134):
+                if scores is not None and i < len(scores) and scores[i] < threshold:
+                    continue
+                x, y = int(keypoints[i][0]), int(keypoints[i][1])
+                if x > eps and y > eps and 0 <= x < W and 0 <= y < H:
+                    self.draw.circle(canvas, (x, y), 4, (0, 0, 255), thickness=-1)
+
+        # Draw face keypoints (24-91) - white dots only, no lines
+        if draw_face and len(keypoints) >= 92:
+            eps = 0.01
+            for i in range(24, 92):
+                if scores is not None and scores[i] < threshold:
+                    continue
+                x, y = int(keypoints[i][0]), int(keypoints[i][1])
+                if x > eps and y > eps and 0 <= x < W and 0 <= y < H:
+                    self.draw.circle(canvas, (x, y), face_point_size, (255, 255, 255), thickness=-1)
+
+        return canvas
+
+class SDPoseDrawKeypoints(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="SDPoseDrawKeypoints",
+            category="image/preprocessors",
+            search_aliases=["openpose", "pose detection", "preprocessor", "keypoints", "pose"],
+            inputs=[
+                io.Custom("POSE_KEYPOINT").Input("keypoints"),
+                io.Boolean.Input("draw_body", default=True),
+                io.Boolean.Input("draw_hands", default=True),
+                io.Boolean.Input("draw_face", default=True),
+                io.Boolean.Input("draw_feet", default=False),
+                io.Int.Input("stick_width", default=4, min=1, max=10, step=1),
+                io.Int.Input("face_point_size", default=3, min=1, max=10, step=1),
+                io.Float.Input("score_threshold", default=0.3, min=0.0, max=1.0, step=0.01),
+            ],
+            outputs=[
+                io.Image.Output(),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, keypoints, draw_body, draw_hands, draw_face, draw_feet, stick_width, face_point_size, score_threshold) -> io.NodeOutput:
+        if not keypoints:
+            return io.NodeOutput(torch.zeros((1, 64, 64, 3), dtype=torch.float32))
+        height = keypoints[0]["canvas_height"]
+        width  = keypoints[0]["canvas_width"]
+
+        def _parse(flat, n):
+            arr = np.array(flat, dtype=np.float32).reshape(n, 3)
+            return arr[:, :2], arr[:, 2]
+
+        def _zeros(n):
+            return np.zeros((n, 2), dtype=np.float32), np.zeros(n, dtype=np.float32)
+
+        pose_outputs = []
+        drawer = KeypointDraw()
+
+        for frame in tqdm(keypoints, desc="Drawing keypoints on frames"):
+            canvas = np.zeros((height, width, 3), dtype=np.uint8)
+            for person in frame["people"]:
+                body_kp,  body_sc  = _parse(person["pose_keypoints_2d"],       18)
+                foot_raw = person.get("foot_keypoints_2d")
+                foot_kp,  foot_sc  = _parse(foot_raw, 6) if foot_raw else _zeros(6)
+                face_kp,  face_sc  = _parse(person["face_keypoints_2d"],       70)
+                face_kp,  face_sc  = face_kp[:68], face_sc[:68]  # drop appended eye kp; body already draws them
+                rhand_kp, rhand_sc = _parse(person["hand_right_keypoints_2d"], 21)
+                lhand_kp, lhand_sc = _parse(person["hand_left_keypoints_2d"],  21)
+
+                kp = np.concatenate([body_kp, foot_kp, face_kp, rhand_kp, lhand_kp], axis=0)
+                sc = np.concatenate([body_sc, foot_sc, face_sc, rhand_sc, lhand_sc], axis=0)
+
+                canvas = drawer.draw_wholebody_keypoints(
+                    canvas, kp, sc,
+                    threshold=score_threshold,
+                    draw_body=draw_body, draw_feet=draw_feet,
+                    draw_face=draw_face, draw_hands=draw_hands,
+                    stick_width=stick_width, face_point_size=face_point_size,
+                )
+            pose_outputs.append(canvas)
+
+        pose_outputs_np = np.stack(pose_outputs) if len(pose_outputs) > 1 else np.expand_dims(pose_outputs[0], 0)
+        final_pose_output = torch.from_numpy(pose_outputs_np).float() / 255.0
+        return io.NodeOutput(final_pose_output)
+
+class SDPoseKeypointExtractor(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="SDPoseKeypointExtractor",
+            category="image/preprocessors",
+            search_aliases=["openpose", "pose detection", "preprocessor", "keypoints", "sdpose"],
+            description="Extract pose keypoints from images using the SDPose model: https://huggingface.co/Comfy-Org/SDPose/tree/main/checkpoints",
+            inputs=[
+                io.Model.Input("model"),
+                io.Vae.Input("vae"),
+                io.Image.Input("image"),
+                io.Int.Input("batch_size", default=16, min=1, max=10000, step=1),
+                io.BoundingBox.Input("bboxes", optional=True, force_input=True, tooltip="Optional bounding boxes for more accurate detections. Required for multi-person detection."),
+            ],
+            outputs=[
+                io.Custom("POSE_KEYPOINT").Output("keypoints", tooltip="Keypoints in OpenPose frame format (canvas_width, canvas_height, people)"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, model, vae, image, batch_size, bboxes=None) -> io.NodeOutput:
+
+        height, width = image.shape[-3], image.shape[-2]
+        context = LotusConditioning().execute().result[0]
+
+        # Use output_block_patch to capture the last 640-channel feature
+        def output_patch(h, hsp, transformer_options):
+            nonlocal captured_feat
+            if h.shape[1] == 640:  # Capture the features for wholebody
+                captured_feat = h.clone()
+            return h, hsp
+
+        model_clone = model.clone()
+        model_clone.model_options["transformer_options"] = {"patches": {"output_block_patch": [output_patch]}}
+
+        if not hasattr(model.model.diffusion_model, 'heatmap_head'):
+            raise ValueError("The provided model does not have a heatmap_head. Please use SDPose model from here https://huggingface.co/Comfy-Org/SDPose/tree/main/checkpoints.")
+
+        head = model.model.diffusion_model.heatmap_head
+        total_images = image.shape[0]
+        captured_feat = None
+
+        model_h = int(head.heatmap_size[0]) * 4   # e.g. 192 * 4 = 768
+        model_w = int(head.heatmap_size[1]) * 4   # e.g. 256 * 4 = 1024
+
+        def _run_on_latent(latent_batch):
+            """Run one forward pass and return (keypoints_list, scores_list) for the batch."""
+            nonlocal captured_feat
+            captured_feat = None
+            _ = comfy.sample.sample(
+                model_clone,
+                noise=torch.zeros_like(latent_batch),
+                steps=1, cfg=1.0,
+                sampler_name="euler", scheduler="simple",
+                positive=context, negative=context,
+                latent_image=latent_batch, disable_noise=True, disable_pbar=True,
+            )
+            return head(captured_feat)  # keypoints_batch, scores_batch
+
+        # all_keypoints / all_scores are lists-of-lists:
+        #   outer index = input image index
+        #   inner index = detected person (one per bbox, or one for full-image)
+        all_keypoints = []  # shape: [n_images][n_persons]
+        all_scores = []     # shape: [n_images][n_persons]
+        pbar = comfy.utils.ProgressBar(total_images)
+
+        if bboxes is not None:
+            if not isinstance(bboxes, list):
+                bboxes = [[bboxes]]
+            elif len(bboxes) == 0:
+                bboxes = [None] * total_images
+            # --- bbox-crop mode: one forward pass per crop -------------------------
+            for img_idx in tqdm(range(total_images), desc="Extracting keypoints from crops"):
+                img = image[img_idx:img_idx + 1]  # (1, H, W, C)
+                # Broadcasting: if fewer bbox lists than images, repeat the last one.
+                img_bboxes = bboxes[min(img_idx, len(bboxes) - 1)] if bboxes else None
+
+                img_keypoints = []
+                img_scores = []
+
+                if img_bboxes:
+                    for bbox in img_bboxes:
+                        x1 = max(0, int(bbox["x"]))
+                        y1 = max(0, int(bbox["y"]))
+                        x2 = min(width,  int(bbox["x"] + bbox["width"]))
+                        y2 = min(height, int(bbox["y"] + bbox["height"]))
+
+                        if x2 <= x1 or y2 <= y1:
+                            continue
+
+                        crop_h_px, crop_w_px = y2 - y1, x2 - x1
+                        crop = img[:, y1:y2, x1:x2, :]  # (1, crop_h, crop_w, C)
+
+                        # scale to fit inside (model_h, model_w) while preserving aspect ratio, then pad to exact model size.
+                        scale = min(model_h / crop_h_px, model_w / crop_w_px)
+                        scaled_h, scaled_w = int(round(crop_h_px * scale)), int(round(crop_w_px * scale))
+                        pad_top, pad_left  = (model_h - scaled_h) // 2, (model_w - scaled_w) // 2
+
+                        crop_chw = crop.permute(0, 3, 1, 2).float()  # BHWC → BCHW
+                        scaled = comfy.utils.common_upscale(crop_chw, scaled_w, scaled_h, upscale_method="bilinear", crop="disabled")
+                        padded = torch.zeros(1, scaled.shape[1], model_h, model_w, dtype=scaled.dtype, device=scaled.device)
+                        padded[:, :, pad_top:pad_top + scaled_h, pad_left:pad_left + scaled_w] = scaled
+                        crop_resized = padded.permute(0, 2, 3, 1)  # BCHW → BHWC
+
+                        latent_crop = vae.encode(crop_resized)
+                        kp_batch, sc_batch = _run_on_latent(latent_crop)
+                        kp, sc = kp_batch[0], sc_batch[0]  # (K, 2), coords in model pixel space
+
+                        # remove padding offset, undo scale, offset to full-image coordinates.
+                        kp = kp.copy() if isinstance(kp, np.ndarray) else np.array(kp, dtype=np.float32)
+                        kp[..., 0] = (kp[..., 0] - pad_left) / scale + x1
+                        kp[..., 1] = (kp[..., 1] - pad_top)  / scale + y1
+
+                        img_keypoints.append(kp)
+                        img_scores.append(sc)
+                else:
+                    # No bboxes for this image – run on the full image
+                    latent_img = vae.encode(img)
+                    kp_batch, sc_batch = _run_on_latent(latent_img)
+                    img_keypoints.append(kp_batch[0])
+                    img_scores.append(sc_batch[0])
+
+                all_keypoints.append(img_keypoints)
+                all_scores.append(img_scores)
+                pbar.update(1)
+
+        else: # full-image mode, batched
+            tqdm_pbar = tqdm(total=total_images, desc="Extracting keypoints")
+            for batch_start in range(0, total_images, batch_size):
+                batch_end = min(batch_start + batch_size, total_images)
+                latent_batch = vae.encode(image[batch_start:batch_end])
+
+                kp_batch, sc_batch = _run_on_latent(latent_batch)
+
+                for kp, sc in zip(kp_batch, sc_batch):
+                    all_keypoints.append([kp])
+                    all_scores.append([sc])
+                    tqdm_pbar.update(1)
+
+                pbar.update(batch_end - batch_start)
+
+        openpose_frames = _to_openpose_frames(all_keypoints, all_scores, height, width)
+        return io.NodeOutput(openpose_frames)
+
+
+def get_face_bboxes(kp2ds, scale, image_shape):
+    h, w = image_shape
+    kp2ds_face = kp2ds.copy()[1:] * (w, h)
+
+    min_x, min_y = np.min(kp2ds_face, axis=0)
+    max_x, max_y = np.max(kp2ds_face, axis=0)
+
+    initial_width = max_x - min_x
+    initial_height = max_y - min_y
+
+    if initial_width <= 0 or initial_height <= 0:
+        return [0, 0, 0, 0]
+
+    initial_area = initial_width * initial_height
+
+    expanded_area = initial_area * scale
+
+    new_width = np.sqrt(expanded_area * (initial_width / initial_height))
+    new_height = np.sqrt(expanded_area * (initial_height / initial_width))
+
+    delta_width = (new_width - initial_width) / 2
+    delta_height = (new_height - initial_height) / 4
+
+    expanded_min_x = max(min_x - delta_width, 0)
+    expanded_max_x = min(max_x + delta_width, w)
+    expanded_min_y = max(min_y - 3 * delta_height, 0)
+    expanded_max_y = min(max_y + delta_height, h)
+
+    return [int(expanded_min_x), int(expanded_max_x), int(expanded_min_y), int(expanded_max_y)]
+
+class SDPoseFaceBBoxes(io.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="SDPoseFaceBBoxes",
+            category="image/preprocessors",
+            search_aliases=["face bbox", "face bounding box", "pose", "keypoints"],
+            inputs=[
+                io.Custom("POSE_KEYPOINT").Input("keypoints"),
+                io.Float.Input("scale", default=1.5, min=1.0, max=10.0, step=0.1, tooltip="Multiplier for the bounding box area around each detected face."),
+                io.Boolean.Input("force_square", default=True, tooltip="Expand the shorter bbox axis so the crop region is always square."),
+            ],
+            outputs=[
+                io.BoundingBox.Output("bboxes", tooltip="Face bounding boxes per frame, compatible with SDPoseKeypointExtractor bboxes input."),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, keypoints, scale, force_square) -> io.NodeOutput:
+        all_bboxes = []
+        for frame in keypoints:
+            h = frame["canvas_height"]
+            w = frame["canvas_width"]
+            frame_bboxes = []
+            for person in frame["people"]:
+                face_flat = person.get("face_keypoints_2d", [])
+                if not face_flat:
+                    continue
+                # Parse absolute-pixel face keypoints (70 kp: 68 landmarks + REye + LEye)
+                face_arr = np.array(face_flat, dtype=np.float32).reshape(-1, 3)
+                face_xy  = face_arr[:, :2]  # (70, 2) in absolute pixels
+
+                kp_norm = face_xy / np.array([w, h], dtype=np.float32)
+                kp_padded = np.vstack([np.zeros((1, 2), dtype=np.float32), kp_norm])  # (71, 2)
+
+                x1, x2, y1, y2 = get_face_bboxes(kp_padded, scale, (h, w))
+                if x2 > x1 and y2 > y1:
+                    if force_square:
+                        bw, bh = x2 - x1, y2 - y1
+                        if bw != bh:
+                            side = max(bw, bh)
+                            cx, cy = (x1 + x2) // 2, (y1 + y2) // 2
+                            half = side // 2
+                            x1 = max(0, cx - half)
+                            y1 = max(0, cy - half)
+                            x2 = min(w, x1 + side)
+                            y2 = min(h, y1 + side)
+                            # Re-anchor if clamped
+                            x1 = max(0, x2 - side)
+                            y1 = max(0, y2 - side)
+                    frame_bboxes.append({"x": x1, "y": y1, "width": x2 - x1, "height": y2 - y1})
+
+            all_bboxes.append(frame_bboxes)
+
+        return io.NodeOutput(all_bboxes)
+
+
+class CropByBBoxes(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="CropByBBoxes",
+            category="image/preprocessors",
+            search_aliases=["crop", "face crop", "bbox crop", "pose", "bounding box"],
+            description="Crop and resize regions from the input image batch based on provided bounding boxes.",
+            inputs=[
+                io.Image.Input("image"),
+                io.BoundingBox.Input("bboxes", force_input=True),
+                io.Int.Input("output_width",  default=512, min=64, max=4096, step=8, tooltip="Width each crop is resized to."),
+                io.Int.Input("output_height", default=512, min=64, max=4096, step=8, tooltip="Height each crop is resized to."),
+                io.Int.Input("padding", default=0, min=0, max=1024, step=1, tooltip="Extra padding in pixels added on each side of the bbox before cropping."),
+            ],
+            outputs=[
+                io.Image.Output(tooltip="All crops stacked into a single image batch."),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, image, bboxes, output_width, output_height, padding) -> io.NodeOutput:
+        total_frames = image.shape[0]
+        img_h = image.shape[1]
+        img_w = image.shape[2]
+        num_ch = image.shape[3]
+
+        if not isinstance(bboxes, list):
+            bboxes = [[bboxes]]
+        elif len(bboxes) == 0:
+            return io.NodeOutput(image)
+
+        crops = []
+
+        for frame_idx in range(total_frames):
+            frame_bboxes = bboxes[min(frame_idx, len(bboxes) - 1)]
+            if not frame_bboxes:
+                continue
+
+            frame_chw = image[frame_idx].permute(2, 0, 1).unsqueeze(0)  # BHWC → BCHW (1, C, H, W)
+
+            # Union all bboxes for this frame into a single crop region
+            x1 = min(int(b["x"]) for b in frame_bboxes)
+            y1 = min(int(b["y"]) for b in frame_bboxes)
+            x2 = max(int(b["x"] + b["width"])  for b in frame_bboxes)
+            y2 = max(int(b["y"] + b["height"]) for b in frame_bboxes)
+
+            if padding > 0:
+                x1 = max(0, x1 - padding)
+                y1 = max(0, y1 - padding)
+                x2 = min(img_w, x2 + padding)
+                y2 = min(img_h, y2 + padding)
+
+            x1, x2 = max(0, x1), min(img_w, x2)
+            y1, y2 = max(0, y1), min(img_h, y2)
+
+            # Fallback for empty/degenerate crops
+            if x2 <= x1 or y2 <= y1:
+                fallback_size = int(min(img_h, img_w) * 0.3)
+                fb_x1 = max(0, (img_w - fallback_size) // 2)
+                fb_y1 = max(0, int(img_h * 0.1))
+                fb_x2 = min(img_w, fb_x1 + fallback_size)
+                fb_y2 = min(img_h, fb_y1 + fallback_size)
+                if fb_x2 <= fb_x1 or fb_y2 <= fb_y1:
+                    crops.append(torch.zeros(1, num_ch, output_height, output_width, dtype=image.dtype, device=image.device))
+                    continue
+                x1, y1, x2, y2 = fb_x1, fb_y1, fb_x2, fb_y2
+
+            crop_chw = frame_chw[:, :, y1:y2, x1:x2]  # (1, C, crop_h, crop_w)
+            resized = comfy.utils.common_upscale(crop_chw, output_width, output_height, upscale_method="bilinear", crop="disabled")
+            crops.append(resized)
+
+        if not crops:
+            return io.NodeOutput(image)
+
+        out_images = torch.cat(crops, dim=0).permute(0, 2, 3, 1)  # (N, H, W, C)
+        return io.NodeOutput(out_images)
+
+
+class SDPoseExtension(ComfyExtension):
+    @override
+    async def get_node_list(self) -> list[type[io.ComfyNode]]:
+        return [
+            SDPoseKeypointExtractor,
+            SDPoseDrawKeypoints,
+            SDPoseFaceBBoxes,
+            CropByBBoxes,
+        ]
+
+async def comfy_entrypoint() -> SDPoseExtension:
+    return SDPoseExtension()

comfy_extras/nodes_video.py

@@ -21,7 +21,7 @@ class SaveWEBM(io.ComfyNode):
             is_experimental=True,
             inputs=[
                 io.Image.Input("images"),
-                io.String.Input("filename_prefix", default="ComfyUI"),
+                io.String.Input("filename_prefix", default="ComfyUI_%year%%month%%day%-%hour%%minute%%second%"),
                 io.Combo.Input("codec", options=["vp9", "av1"]),
                 io.Float.Input("fps", default=24.0, min=0.01, max=1000.0, step=0.01),
                 io.Float.Input("crf", default=32.0, min=0, max=63.0, step=1, tooltip="Higher crf means lower quality with a smaller file size, lower crf means higher quality higher filesize."),
@@ -77,7 +77,7 @@ class SaveVideo(io.ComfyNode):
             description="Saves the input images to your ComfyUI output directory.",
             inputs=[
                 io.Video.Input("video", tooltip="The video to save."),
-                io.String.Input("filename_prefix", default="video/ComfyUI", tooltip="The prefix for the file to save. This may include formatting information such as %date:yyyy-MM-dd% or %Empty Latent Image.width% to include values from nodes."),
+                io.String.Input("filename_prefix", default="video/ComfyUI_%year%%month%%day%-%hour%%minute%%second%", tooltip="The prefix for the file to save. This may include formatting information such as %date:yyyy-MM-dd% or %Empty Latent Image.width% to include values from nodes."),
                 io.Combo.Input("format", options=Types.VideoContainer.as_input(), default="auto", tooltip="The format to save the video as."),
                 io.Combo.Input("codec", options=Types.VideoCodec.as_input(), default="auto", tooltip="The codec to use for the video."),
             ],

nodes.py

@@ -976,7 +976,7 @@ class CLIPLoader:
     @classmethod
     def INPUT_TYPES(s):
         return {"required": { "clip_name": (folder_paths.get_filename_list("text_encoders"), ),
-                              "type": (["stable_diffusion", "stable_cascade", "sd3", "stable_audio", "mochi", "ltxv", "pixart", "cosmos", "lumina2", "wan", "hidream", "chroma", "ace", "omnigen2", "qwen_image", "hunyuan_image", "flux2", "ovis"], ),
+                              "type": (["stable_diffusion", "stable_cascade", "sd3", "stable_audio", "mochi", "ltxv", "pixart", "cosmos", "lumina2", "wan", "hidream", "chroma", "ace", "omnigen2", "qwen_image", "hunyuan_image", "flux2", "ovis", "longcat_image"], ),
                               },
                 "optional": {
                               "device": (["default", "cpu"], {"advanced": True}),
@@ -1638,7 +1638,7 @@ class SaveImage:
         return {
             "required": {
                 "images": ("IMAGE", {"tooltip": "The images to save."}),
-                "filename_prefix": ("STRING", {"default": "ComfyUI", "tooltip": "The prefix for the file to save. This may include formatting information such as %date:yyyy-MM-dd% or %Empty Latent Image.width% to include values from nodes."})
+                "filename_prefix": ("STRING", {"default": "ComfyUI_%year%%month%%day%-%hour%%minute%%second%", "tooltip": "The prefix for the file to save. This may include formatting information such as %date:yyyy-MM-dd% or %Empty Latent Image.width% to include values from nodes."})
             },
             "hidden": {
                 "prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"
@@ -2435,6 +2435,7 @@ async def init_builtin_extra_nodes():
         "nodes_audio_encoder.py",
         "nodes_rope.py",
         "nodes_logic.py",
+        "nodes_resolution.py",
         "nodes_nop.py",
         "nodes_kandinsky5.py",
         "nodes_wanmove.py",
@@ -2447,6 +2448,7 @@ async def init_builtin_extra_nodes():
         "nodes_toolkit.py",
         "nodes_replacements.py",
         "nodes_nag.py",
+        "nodes_sdpose.py",
     ]
 
     import_failed = []

requirements.txt

@@ -31,5 +31,4 @@ spandrel
 pydantic~=2.0
 pydantic-settings~=2.0
 PyOpenGL
-PyOpenGL-accelerate
 glfw

tests-unit/comfy_test/model_detection_test.py

@@ -0,0 +1,112 @@
+import torch
+
+from comfy.model_detection import detect_unet_config, model_config_from_unet_config
+import comfy.supported_models
+
+
+def _make_longcat_comfyui_sd():
+    """Minimal ComfyUI-format state dict for pre-converted LongCat-Image weights."""
+    sd = {}
+    H = 32  # Reduce hidden state dimension to reduce memory usage
+    C_IN = 16
+    C_CTX = 3584
+
+    sd["img_in.weight"] = torch.empty(H, C_IN * 4)
+    sd["img_in.bias"] = torch.empty(H)
+    sd["txt_in.weight"] = torch.empty(H, C_CTX)
+    sd["txt_in.bias"] = torch.empty(H)
+
+    sd["time_in.in_layer.weight"] = torch.empty(H, 256)
+    sd["time_in.in_layer.bias"] = torch.empty(H)
+    sd["time_in.out_layer.weight"] = torch.empty(H, H)
+    sd["time_in.out_layer.bias"] = torch.empty(H)
+
+    sd["final_layer.adaLN_modulation.1.weight"] = torch.empty(2 * H, H)
+    sd["final_layer.adaLN_modulation.1.bias"] = torch.empty(2 * H)
+    sd["final_layer.linear.weight"] = torch.empty(C_IN * 4, H)
+    sd["final_layer.linear.bias"] = torch.empty(C_IN * 4)
+
+    for i in range(19):
+        sd[f"double_blocks.{i}.img_attn.norm.key_norm.weight"] = torch.empty(128)
+        sd[f"double_blocks.{i}.img_attn.qkv.weight"] = torch.empty(3 * H, H)
+        sd[f"double_blocks.{i}.img_mod.lin.weight"] = torch.empty(H, H)
+    for i in range(38):
+        sd[f"single_blocks.{i}.modulation.lin.weight"] = torch.empty(H, H)
+
+    return sd
+
+
+def _make_flux_schnell_comfyui_sd():
+    """Minimal ComfyUI-format state dict for standard Flux Schnell."""
+    sd = {}
+    H = 32  # Reduce hidden state dimension to reduce memory usage
+    C_IN = 16
+
+    sd["img_in.weight"] = torch.empty(H, C_IN * 4)
+    sd["img_in.bias"] = torch.empty(H)
+    sd["txt_in.weight"] = torch.empty(H, 4096)
+    sd["txt_in.bias"] = torch.empty(H)
+
+    sd["double_blocks.0.img_attn.norm.key_norm.weight"] = torch.empty(128)
+    sd["double_blocks.0.img_attn.qkv.weight"] = torch.empty(3 * H, H)
+    sd["double_blocks.0.img_mod.lin.weight"] = torch.empty(H, H)
+
+    for i in range(19):
+        sd[f"double_blocks.{i}.img_attn.norm.key_norm.weight"] = torch.empty(128)
+    for i in range(38):
+        sd[f"single_blocks.{i}.modulation.lin.weight"] = torch.empty(H, H)
+
+    return sd
+
+
+class TestModelDetection:
+    """Verify that first-match model detection selects the correct model
+    based on list ordering and unet_config specificity."""
+
+    def test_longcat_before_schnell_in_models_list(self):
+        """LongCatImage must appear before FluxSchnell in the models list."""
+        models = comfy.supported_models.models
+        longcat_idx = next(i for i, m in enumerate(models) if m.__name__ == "LongCatImage")
+        schnell_idx = next(i for i, m in enumerate(models) if m.__name__ == "FluxSchnell")
+        assert longcat_idx < schnell_idx, (
+            f"LongCatImage (index {longcat_idx}) must come before "
+            f"FluxSchnell (index {schnell_idx}) in the models list"
+        )
+
+    def test_longcat_comfyui_detected_as_longcat(self):
+        sd = _make_longcat_comfyui_sd()
+        unet_config = detect_unet_config(sd, "")
+        assert unet_config is not None
+        assert unet_config["image_model"] == "flux"
+        assert unet_config["context_in_dim"] == 3584
+        assert unet_config["vec_in_dim"] is None
+        assert unet_config["guidance_embed"] is False
+        assert unet_config["txt_ids_dims"] == [1, 2]
+
+        model_config = model_config_from_unet_config(unet_config, sd)
+        assert model_config is not None
+        assert type(model_config).__name__ == "LongCatImage"
+
+    def test_longcat_comfyui_keys_pass_through_unchanged(self):
+        """Pre-converted weights should not be transformed by process_unet_state_dict."""
+        sd = _make_longcat_comfyui_sd()
+        unet_config = detect_unet_config(sd, "")
+        model_config = model_config_from_unet_config(unet_config, sd)
+
+        processed = model_config.process_unet_state_dict(dict(sd))
+        assert "img_in.weight" in processed
+        assert "txt_in.weight" in processed
+        assert "time_in.in_layer.weight" in processed
+        assert "final_layer.linear.weight" in processed
+
+    def test_flux_schnell_comfyui_detected_as_flux_schnell(self):
+        sd = _make_flux_schnell_comfyui_sd()
+        unet_config = detect_unet_config(sd, "")
+        assert unet_config is not None
+        assert unet_config["image_model"] == "flux"
+        assert unet_config["context_in_dim"] == 4096
+        assert unet_config["txt_ids_dims"] == []
+
+        model_config = model_config_from_unet_config(unet_config, sd)
+        assert model_config is not None
+        assert type(model_config).__name__ == "FluxSchnell"