Merge f334f2db3d into 3e3ed8cc2a

Co-authored-by: Jedrzej Kosinski <kosinkadink1@gmail.com>

.ci/windows_amd_base_files/run_amd_gpu_enable_dynamic_vram.bat

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

.gitignore

@@ -8,6 +8,7 @@ __pycache__/
 /custom_nodes/
 !custom_nodes/example_node.py.example
 extra_model_paths.yaml
+extra_paths.yaml
 /.vs
 .vscode/
 .idea/

CODEOWNERS

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

README.md

@@ -193,13 +193,15 @@ If you have trouble extracting it, right click the file -> properties -> unblock
 
 The portable above currently comes with python 3.13 and pytorch cuda 13.0. Update your Nvidia drivers if it doesn't start.
 
-#### Alternative Downloads:
+#### All Official Portable Downloads:
 
 [Portable for AMD GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_amd.7z)
 
-[Experimental portable for Intel GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_intel.7z)
+[Portable for Intel GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_intel.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).
+[Portable for Nvidia GPUs](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia.7z) (supports 20 series and above).
+
+[Portable for Nvidia GPUs with pytorch cuda 12.6 and python 3.12](https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia_cu126.7z) (Supports Nvidia 10 series and older GPUs).
 
 #### How do I share models between another UI and ComfyUI?
 

comfy/cli_args.py

@@ -90,7 +90,6 @@ parser.add_argument("--force-channels-last", action="store_true", help="Force ch
 parser.add_argument("--directml", type=int, nargs="?", metavar="DIRECTML_DEVICE", const=-1, help="Use torch-directml.")
 
 parser.add_argument("--oneapi-device-selector", type=str, default=None, metavar="SELECTOR_STRING", help="Sets the oneAPI device(s) this instance will use.")
-parser.add_argument("--disable-ipex-optimize", action="store_true", help="Disables ipex.optimize default when loading models with Intel's Extension for Pytorch.")
 parser.add_argument("--supports-fp8-compute", action="store_true", help="ComfyUI will act like if the device supports fp8 compute.")
 
 class LatentPreviewMethod(enum.Enum):

comfy/model_management.py

@@ -112,10 +112,6 @@ if args.directml is not None:
     # torch_directml.disable_tiled_resources(True)
     lowvram_available = False #TODO: need to find a way to get free memory in directml before this can be enabled by default.
 
-try:
-    import intel_extension_for_pytorch as ipex  # noqa: F401
-except:
-    pass
 
 try:
     _ = torch.xpu.device_count()
@@ -583,9 +579,6 @@ class LoadedModel:
 
         real_model = self.model.model
 
-        if is_intel_xpu() and not args.disable_ipex_optimize and 'ipex' in globals() and real_model is not None:
-            with torch.no_grad():
-                real_model = ipex.optimize(real_model.eval(), inplace=True, graph_mode=True, concat_linear=True)
 
         self.real_model = weakref.ref(real_model)
         self.model_finalizer = weakref.finalize(real_model, cleanup_models)
@@ -1581,10 +1574,7 @@ def should_use_fp16(device=None, model_params=0, prioritize_performance=True, ma
         return False
 
     if is_intel_xpu():
-        if torch_version_numeric < (2, 3):
-            return True
-        else:
-            return torch.xpu.get_device_properties(device).has_fp16
+        return torch.xpu.get_device_properties(device).has_fp16
 
     if is_ascend_npu():
         return True
@@ -1650,10 +1640,7 @@ def should_use_bf16(device=None, model_params=0, prioritize_performance=True, ma
         return False
 
     if is_intel_xpu():
-        if torch_version_numeric < (2, 3):
-            return True
-        else:
-            return torch.xpu.is_bf16_supported()
+        return torch.xpu.is_bf16_supported()
 
     if is_ascend_npu():
         return True
@@ -1784,6 +1771,7 @@ def soft_empty_cache(force=False):
     if cpu_state == CPUState.MPS:
         torch.mps.empty_cache()
     elif is_intel_xpu():
+        torch.xpu.synchronize()
         torch.xpu.empty_cache()
     elif is_ascend_npu():
         torch.npu.empty_cache()

comfy/supported_models.py

@@ -1879,6 +1879,86 @@ class CogVideoX_I2V(CogVideoX_T2V):
         out = model_base.CogVideoX(self, image_to_video=True, device=device)
         return out
 
-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, ZImagePixelSpace, ZImage, Lumina2, WAN22_T2V, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, WAN21_Vace, WAN21_Camera, WAN22_Camera, WAN22_S2V, WAN21_HuMo, WAN22_Animate, WAN21_FlowRVS, WAN21_SCAIL, Hunyuan3Dv2mini, Hunyuan3Dv2, Hunyuan3Dv2_1, HiDream, Chroma, ChromaRadiance, ACEStep, ACEStep15, Omnigen2, QwenImage, Flux2, Kandinsky5Image, Kandinsky5, Anima, RT_DETR_v4, ErnieImage, SAM3, SAM31, CogVideoX_I2V, CogVideoX_T2V]
 
-models += [SVD_img2vid]
+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,
+    ZImagePixelSpace,
+    ZImage,
+    Lumina2,
+    WAN22_T2V,
+    WAN21_T2V,
+    WAN21_I2V,
+    WAN21_FunControl2V,
+    WAN21_Vace,
+    WAN21_Camera,
+    WAN22_Camera,
+    WAN22_S2V,
+    WAN21_HuMo,
+    WAN22_Animate,
+    WAN21_FlowRVS,
+    WAN21_SCAIL,
+    Hunyuan3Dv2mini,
+    Hunyuan3Dv2,
+    Hunyuan3Dv2_1,
+    HiDream,
+    Chroma,
+    ChromaRadiance,
+    ACEStep,
+    ACEStep15,
+    Omnigen2,
+    QwenImage,
+    Flux2,
+    Kandinsky5Image,
+    Kandinsky5,
+    Anima,
+    RT_DETR_v4,
+    ErnieImage,
+    SAM3,
+    SAM31,
+    CogVideoX_I2V,
+    CogVideoX_T2V,
+    SVD_img2vid,
+]

comfy_api_nodes/apis/moonvalley.py

@@ -1,152 +0,0 @@
-from enum import Enum
-from typing import Optional, Dict, Any
-
-from pydantic import BaseModel, Field, StrictBytes
-
-
-class MoonvalleyPromptResponse(BaseModel):
-    error: Optional[Dict[str, Any]] = None
-    frame_conditioning: Optional[Dict[str, Any]] = None
-    id: Optional[str] = None
-    inference_params: Optional[Dict[str, Any]] = None
-    meta: Optional[Dict[str, Any]] = None
-    model_params: Optional[Dict[str, Any]] = None
-    output_url: Optional[str] = None
-    prompt_text: Optional[str] = None
-    status: Optional[str] = None
-
-
-class MoonvalleyTextToVideoInferenceParams(BaseModel):
-    add_quality_guidance: Optional[bool] = Field(
-        True, description='Whether to add quality guidance'
-    )
-    caching_coefficient: Optional[float] = Field(
-        0.3, description='Caching coefficient for optimization'
-    )
-    caching_cooldown: Optional[int] = Field(
-        3, description='Number of caching cooldown steps'
-    )
-    caching_warmup: Optional[int] = Field(
-        3, description='Number of caching warmup steps'
-    )
-    clip_value: Optional[float] = Field(
-        3, description='CLIP value for generation control'
-    )
-    conditioning_frame_index: Optional[int] = Field(
-        0, description='Index of the conditioning frame'
-    )
-    cooldown_steps: Optional[int] = Field(
-        75, description='Number of cooldown steps (calculated based on num_frames)'
-    )
-    fps: Optional[int] = Field(
-        24, description='Frames per second of the generated video'
-    )
-    guidance_scale: Optional[float] = Field(
-        10, description='Guidance scale for generation control'
-    )
-    height: Optional[int] = Field(
-        1080, description='Height of the generated video in pixels'
-    )
-    negative_prompt: Optional[str] = Field(None, description='Negative prompt text')
-    num_frames: Optional[int] = Field(64, description='Number of frames to generate')
-    seed: Optional[int] = Field(
-        None, description='Random seed for generation (default: random)'
-    )
-    shift_value: Optional[float] = Field(
-        3, description='Shift value for generation control'
-    )
-    steps: Optional[int] = Field(80, description='Number of denoising steps')
-    use_guidance_schedule: Optional[bool] = Field(
-        True, description='Whether to use guidance scheduling'
-    )
-    use_negative_prompts: Optional[bool] = Field(
-        False, description='Whether to use negative prompts'
-    )
-    use_timestep_transform: Optional[bool] = Field(
-        True, description='Whether to use timestep transformation'
-    )
-    warmup_steps: Optional[int] = Field(
-        0, description='Number of warmup steps (calculated based on num_frames)'
-    )
-    width: Optional[int] = Field(
-        1920, description='Width of the generated video in pixels'
-    )
-
-
-class MoonvalleyTextToVideoRequest(BaseModel):
-    image_url: Optional[str] = None
-    inference_params: Optional[MoonvalleyTextToVideoInferenceParams] = None
-    prompt_text: Optional[str] = None
-    webhook_url: Optional[str] = None
-
-
-class MoonvalleyUploadFileRequest(BaseModel):
-    file: Optional[StrictBytes] = None
-
-
-class MoonvalleyUploadFileResponse(BaseModel):
-    access_url: Optional[str] = None
-
-
-class MoonvalleyVideoToVideoInferenceParams(BaseModel):
-    add_quality_guidance: Optional[bool] = Field(
-        True, description='Whether to add quality guidance'
-    )
-    caching_coefficient: Optional[float] = Field(
-        0.3, description='Caching coefficient for optimization'
-    )
-    caching_cooldown: Optional[int] = Field(
-        3, description='Number of caching cooldown steps'
-    )
-    caching_warmup: Optional[int] = Field(
-        3, description='Number of caching warmup steps'
-    )
-    clip_value: Optional[float] = Field(
-        3, description='CLIP value for generation control'
-    )
-    conditioning_frame_index: Optional[int] = Field(
-        0, description='Index of the conditioning frame'
-    )
-    cooldown_steps: Optional[int] = Field(
-        36, description='Number of cooldown steps (calculated based on num_frames)'
-    )
-    guidance_scale: Optional[float] = Field(
-        15, description='Guidance scale for generation control'
-    )
-    negative_prompt: Optional[str] = Field(None, description='Negative prompt text')
-    seed: Optional[int] = Field(
-        None, description='Random seed for generation (default: random)'
-    )
-    shift_value: Optional[float] = Field(
-        3, description='Shift value for generation control'
-    )
-    steps: Optional[int] = Field(80, description='Number of denoising steps')
-    use_guidance_schedule: Optional[bool] = Field(
-        True, description='Whether to use guidance scheduling'
-    )
-    use_negative_prompts: Optional[bool] = Field(
-        False, description='Whether to use negative prompts'
-    )
-    use_timestep_transform: Optional[bool] = Field(
-        True, description='Whether to use timestep transformation'
-    )
-    warmup_steps: Optional[int] = Field(
-        24, description='Number of warmup steps (calculated based on num_frames)'
-    )
-
-
-class ControlType(str, Enum):
-    motion_control = 'motion_control'
-    pose_control = 'pose_control'
-
-
-class MoonvalleyVideoToVideoRequest(BaseModel):
-    control_type: ControlType = Field(
-        ..., description='Supported types for video control'
-    )
-    inference_params: Optional[MoonvalleyVideoToVideoInferenceParams] = None
-    prompt_text: str = Field(..., description='Describes the video to generate')
-    video_url: str = Field(..., description='Url to control video')
-    webhook_url: Optional[str] = Field(
-        None, description='Optional webhook URL for notifications'
-    )

comfy_api_nodes/nodes_bytedance.py

@@ -1403,7 +1403,6 @@ class ByteDance2TextToVideoNode(IO.ComfyNode):
             status_extractor=lambda r: r.status,
             price_extractor=_seedance2_price_extractor(model_id, has_video_input=False),
             poll_interval=9,
-            max_poll_attempts=180,
         )
         return IO.NodeOutput(await download_url_to_video_output(response.content.video_url))
 
@@ -1585,7 +1584,6 @@ class ByteDance2FirstLastFrameNode(IO.ComfyNode):
             status_extractor=lambda r: r.status,
             price_extractor=_seedance2_price_extractor(model_id, has_video_input=False),
             poll_interval=9,
-            max_poll_attempts=180,
         )
         return IO.NodeOutput(await download_url_to_video_output(response.content.video_url))
 
@@ -1907,7 +1905,6 @@ class ByteDance2ReferenceNode(IO.ComfyNode):
             status_extractor=lambda r: r.status,
             price_extractor=_seedance2_price_extractor(model_id, has_video_input=has_video_input),
             poll_interval=9,
-            max_poll_attempts=180,
         )
         return IO.NodeOutput(await download_url_to_video_output(response.content.video_url))
 

comfy_api_nodes/nodes_hitpaw.py

@@ -178,7 +178,6 @@ class HitPawGeneralImageEnhance(IO.ComfyNode):
             status_extractor=lambda x: x.data.status,
             price_extractor=lambda x: request_price,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_image_tensor(final_response.data.res_url))
 
@@ -324,7 +323,6 @@ class HitPawVideoEnhance(IO.ComfyNode):
             status_extractor=lambda x: x.data.status,
             price_extractor=lambda x: request_price,
             poll_interval=10.0,
-            max_poll_attempts=320,
         )
         return IO.NodeOutput(await download_url_to_video_output(final_response.data.res_url))
 

comfy_api_nodes/nodes_kling.py

@@ -276,7 +276,6 @@ async def finish_omni_video_task(cls: type[IO.ComfyNode], response: TaskStatusRe
         cls,
         ApiEndpoint(path=f"/proxy/kling/v1/videos/omni-video/{response.data.task_id}"),
         response_model=TaskStatusResponse,
-        max_poll_attempts=280,
         status_extractor=lambda r: (r.data.task_status if r.data else None),
     )
     return IO.NodeOutput(await download_url_to_video_output(final_response.data.task_result.videos[0].url))
@@ -3062,7 +3061,6 @@ class KlingVideoNode(IO.ComfyNode):
             cls,
             ApiEndpoint(path=poll_path),
             response_model=TaskStatusResponse,
-            max_poll_attempts=280,
             status_extractor=lambda r: (r.data.task_status if r.data else None),
         )
         return IO.NodeOutput(await download_url_to_video_output(final_response.data.task_result.videos[0].url))
@@ -3188,7 +3186,6 @@ class KlingFirstLastFrameNode(IO.ComfyNode):
             cls,
             ApiEndpoint(path=f"/proxy/kling/v1/videos/image2video/{response.data.task_id}"),
             response_model=TaskStatusResponse,
-            max_poll_attempts=280,
             status_extractor=lambda r: (r.data.task_status if r.data else None),
         )
         return IO.NodeOutput(await download_url_to_video_output(final_response.data.task_result.videos[0].url))

comfy_api_nodes/nodes_magnific.py

@@ -230,7 +230,6 @@ class MagnificImageUpscalerCreativeNode(IO.ComfyNode):
             status_extractor=lambda x: x.status,
             price_extractor=lambda _: price_usd,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_image_tensor(final_response.generated[0]))
 
@@ -391,7 +390,6 @@ class MagnificImageUpscalerPreciseV2Node(IO.ComfyNode):
             status_extractor=lambda x: x.status,
             price_extractor=lambda _: price_usd,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_image_tensor(final_response.generated[0]))
 
@@ -541,7 +539,6 @@ class MagnificImageStyleTransferNode(IO.ComfyNode):
             response_model=TaskResponse,
             status_extractor=lambda x: x.status,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_image_tensor(final_response.generated[0]))
 
@@ -782,7 +779,6 @@ class MagnificImageRelightNode(IO.ComfyNode):
             response_model=TaskResponse,
             status_extractor=lambda x: x.status,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_image_tensor(final_response.generated[0]))
 
@@ -924,7 +920,6 @@ class MagnificImageSkinEnhancerNode(IO.ComfyNode):
             response_model=TaskResponse,
             status_extractor=lambda x: x.status,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_image_tensor(final_response.generated[0]))
 

comfy_api_nodes/nodes_moonvalley.py

@@ -1,534 +0,0 @@
-import logging
-
-from typing_extensions import override
-
-from comfy_api.latest import IO, ComfyExtension, Input
-from comfy_api_nodes.apis.moonvalley import (
-    MoonvalleyPromptResponse,
-    MoonvalleyTextToVideoInferenceParams,
-    MoonvalleyTextToVideoRequest,
-    MoonvalleyVideoToVideoInferenceParams,
-    MoonvalleyVideoToVideoRequest,
-)
-from comfy_api_nodes.util import (
-    ApiEndpoint,
-    download_url_to_video_output,
-    poll_op,
-    sync_op,
-    trim_video,
-    upload_images_to_comfyapi,
-    upload_video_to_comfyapi,
-    validate_container_format_is_mp4,
-    validate_image_dimensions,
-    validate_string,
-)
-
-API_UPLOADS_ENDPOINT = "/proxy/moonvalley/uploads"
-API_PROMPTS_ENDPOINT = "/proxy/moonvalley/prompts"
-API_VIDEO2VIDEO_ENDPOINT = "/proxy/moonvalley/prompts/video-to-video"
-API_TXT2VIDEO_ENDPOINT = "/proxy/moonvalley/prompts/text-to-video"
-API_IMG2VIDEO_ENDPOINT = "/proxy/moonvalley/prompts/image-to-video"
-
-MIN_WIDTH = 300
-MIN_HEIGHT = 300
-
-MAX_WIDTH = 10000
-MAX_HEIGHT = 10000
-
-MIN_VID_WIDTH = 300
-MIN_VID_HEIGHT = 300
-
-MAX_VID_WIDTH = 10000
-MAX_VID_HEIGHT = 10000
-
-MAX_VIDEO_SIZE = 1024 * 1024 * 1024  # 1 GB max for in-memory video processing
-
-MOONVALLEY_MAREY_MAX_PROMPT_LENGTH = 5000
-
-
-def is_valid_task_creation_response(response: MoonvalleyPromptResponse) -> bool:
-    """Verifies that the initial response contains a task ID."""
-    return bool(response.id)
-
-
-def validate_task_creation_response(response) -> None:
-    if not is_valid_task_creation_response(response):
-        error_msg = f"Moonvalley Marey API: Initial request failed. Code: {response.code}, Message: {response.message}, Data: {response}"
-        logging.error(error_msg)
-        raise RuntimeError(error_msg)
-
-
-def validate_video_to_video_input(video: Input.Video) -> Input.Video:
-    """
-    Validates and processes video input for Moonvalley Video-to-Video generation.
-
-    Args:
-        video: Input video to validate
-
-    Returns:
-        Validated and potentially trimmed video
-
-    Raises:
-        ValueError: If video doesn't meet requirements
-        MoonvalleyApiError: If video duration is too short
-    """
-    width, height = _get_video_dimensions(video)
-    _validate_video_dimensions(width, height)
-    validate_container_format_is_mp4(video)
-
-    return _validate_and_trim_duration(video)
-
-
-def _get_video_dimensions(video: Input.Video) -> tuple[int, int]:
-    """Extracts video dimensions with error handling."""
-    try:
-        return video.get_dimensions()
-    except Exception as e:
-        logging.error("Error getting dimensions of video: %s", e)
-        raise ValueError(f"Cannot get video dimensions: {e}") from e
-
-
-def _validate_video_dimensions(width: int, height: int) -> None:
-    """Validates video dimensions meet Moonvalley V2V requirements."""
-    supported_resolutions = {
-        (1920, 1080),
-        (1080, 1920),
-        (1152, 1152),
-        (1536, 1152),
-        (1152, 1536),
-    }
-
-    if (width, height) not in supported_resolutions:
-        supported_list = ", ".join([f"{w}x{h}" for w, h in sorted(supported_resolutions)])
-        raise ValueError(f"Resolution {width}x{height} not supported. Supported: {supported_list}")
-
-
-def _validate_and_trim_duration(video: Input.Video) -> Input.Video:
-    """Validates video duration and trims to 5 seconds if needed."""
-    duration = video.get_duration()
-    _validate_minimum_duration(duration)
-    return _trim_if_too_long(video, duration)
-
-
-def _validate_minimum_duration(duration: float) -> None:
-    """Ensures video is at least 5 seconds long."""
-    if duration < 5:
-        raise ValueError("Input video must be at least 5 seconds long.")
-
-
-def _trim_if_too_long(video: Input.Video, duration: float) -> Input.Video:
-    """Trims video to 5 seconds if longer."""
-    if duration > 5:
-        return trim_video(video, 5)
-    return video
-
-
-def parse_width_height_from_res(resolution: str):
-    # Accepts a string like "16:9 (1920 x 1080)" and returns width, height as a dict
-    res_map = {
-        "16:9 (1920 x 1080)": {"width": 1920, "height": 1080},
-        "9:16 (1080 x 1920)": {"width": 1080, "height": 1920},
-        "1:1 (1152 x 1152)": {"width": 1152, "height": 1152},
-        "4:3 (1536 x 1152)": {"width": 1536, "height": 1152},
-        "3:4 (1152 x 1536)": {"width": 1152, "height": 1536},
-        # "21:9 (2560 x 1080)": {"width": 2560, "height": 1080},
-    }
-    return res_map.get(resolution, {"width": 1920, "height": 1080})
-
-
-def parse_control_parameter(value):
-    control_map = {
-        "Motion Transfer": "motion_control",
-        "Canny": "canny_control",
-        "Pose Transfer": "pose_control",
-        "Depth": "depth_control",
-    }
-    return control_map.get(value, control_map["Motion Transfer"])
-
-
-async def get_response(cls: type[IO.ComfyNode], task_id: str) -> MoonvalleyPromptResponse:
-    return await poll_op(
-        cls,
-        ApiEndpoint(path=f"{API_PROMPTS_ENDPOINT}/{task_id}"),
-        response_model=MoonvalleyPromptResponse,
-        status_extractor=lambda r: (r.status if r and r.status else None),
-        poll_interval=16.0,
-        max_poll_attempts=240,
-    )
-
-
-class MoonvalleyImg2VideoNode(IO.ComfyNode):
-
-    @classmethod
-    def define_schema(cls) -> IO.Schema:
-        return IO.Schema(
-            node_id="MoonvalleyImg2VideoNode",
-            display_name="Moonvalley Marey Image to Video",
-            category="api node/video/Moonvalley Marey",
-            description="Moonvalley Marey Image to Video Node",
-            inputs=[
-                IO.Image.Input(
-                    "image",
-                    tooltip="The reference image used to generate the video",
-                ),
-                IO.String.Input(
-                    "prompt",
-                    multiline=True,
-                ),
-                IO.String.Input(
-                    "negative_prompt",
-                    multiline=True,
-                    default="<synthetic> <scene cut> gopro, bright, contrast, static, overexposed, vignette, "
-                    "artifacts, still, noise, texture, scanlines, videogame, 360 camera, VR, transition, "
-                    "flare, saturation, distorted, warped, wide angle, saturated, vibrant, glowing, "
-                    "cross dissolve, cheesy, ugly hands, mutated hands, mutant, disfigured, extra fingers, "
-                    "blown out, horrible, blurry, worst quality, bad, dissolve, melt, fade in, fade out, "
-                    "wobbly, weird, low quality, plastic, stock footage, video camera, boring",
-                    tooltip="Negative prompt text",
-                ),
-                IO.Combo.Input(
-                    "resolution",
-                    options=[
-                        "16:9 (1920 x 1080)",
-                        "9:16 (1080 x 1920)",
-                        "1:1 (1152 x 1152)",
-                        "4:3 (1536 x 1152)",
-                        "3:4 (1152 x 1536)",
-                        # "21:9 (2560 x 1080)",
-                    ],
-                    default="16:9 (1920 x 1080)",
-                    tooltip="Resolution of the output video",
-                ),
-                IO.Float.Input(
-                    "prompt_adherence",
-                    default=4.5,
-                    min=1.0,
-                    max=20.0,
-                    step=1.0,
-                    tooltip="Guidance scale for generation control",
-                ),
-                IO.Int.Input(
-                    "seed",
-                    default=9,
-                    min=0,
-                    max=4294967295,
-                    step=1,
-                    display_mode=IO.NumberDisplay.number,
-                    tooltip="Random seed value",
-                    control_after_generate=True,
-                ),
-                IO.Int.Input(
-                    "steps",
-                    default=80,
-                    min=75,  # steps should be greater or equal to cooldown_steps(75) + warmup_steps(0)
-                    max=100,
-                    step=1,
-                    tooltip="Number of denoising steps",
-                ),
-            ],
-            outputs=[IO.Video.Output()],
-            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(
-                depends_on=IO.PriceBadgeDepends(),
-                expr="""{"type":"usd","usd": 1.5}""",
-            ),
-        )
-
-    @classmethod
-    async def execute(
-        cls,
-        image: Input.Image,
-        prompt: str,
-        negative_prompt: str,
-        resolution: str,
-        prompt_adherence: float,
-        seed: int,
-        steps: int,
-    ) -> IO.NodeOutput:
-        validate_image_dimensions(image, min_width=300, min_height=300, max_height=MAX_HEIGHT, max_width=MAX_WIDTH)
-        validate_string(prompt, min_length=1, max_length=MOONVALLEY_MAREY_MAX_PROMPT_LENGTH)
-        validate_string(negative_prompt, field_name="negative_prompt", max_length=MOONVALLEY_MAREY_MAX_PROMPT_LENGTH)
-        width_height = parse_width_height_from_res(resolution)
-
-        inference_params = MoonvalleyTextToVideoInferenceParams(
-            negative_prompt=negative_prompt,
-            steps=steps,
-            seed=seed,
-            guidance_scale=prompt_adherence,
-            width=width_height["width"],
-            height=width_height["height"],
-            use_negative_prompts=True,
-        )
-
-        # Get MIME type from tensor - assuming PNG format for image tensors
-        mime_type = "image/png"
-        image_url = (await upload_images_to_comfyapi(cls, image, max_images=1, mime_type=mime_type))[0]
-        task_creation_response = await sync_op(
-            cls,
-            endpoint=ApiEndpoint(path=API_IMG2VIDEO_ENDPOINT, method="POST"),
-            response_model=MoonvalleyPromptResponse,
-            data=MoonvalleyTextToVideoRequest(
-                image_url=image_url, prompt_text=prompt, inference_params=inference_params
-            ),
-        )
-        validate_task_creation_response(task_creation_response)
-        final_response = await get_response(cls, task_creation_response.id)
-        video = await download_url_to_video_output(final_response.output_url)
-        return IO.NodeOutput(video)
-
-
-class MoonvalleyVideo2VideoNode(IO.ComfyNode):
-
-    @classmethod
-    def define_schema(cls) -> IO.Schema:
-        return IO.Schema(
-            node_id="MoonvalleyVideo2VideoNode",
-            display_name="Moonvalley Marey Video to Video",
-            category="api node/video/Moonvalley Marey",
-            description="",
-            inputs=[
-                IO.String.Input(
-                    "prompt",
-                    multiline=True,
-                    tooltip="Describes the video to generate",
-                ),
-                IO.String.Input(
-                    "negative_prompt",
-                    multiline=True,
-                    default="<synthetic> <scene cut> gopro, bright, contrast, static, overexposed, vignette, "
-                    "artifacts, still, noise, texture, scanlines, videogame, 360 camera, VR, transition, "
-                    "flare, saturation, distorted, warped, wide angle, saturated, vibrant, glowing, "
-                    "cross dissolve, cheesy, ugly hands, mutated hands, mutant, disfigured, extra fingers, "
-                    "blown out, horrible, blurry, worst quality, bad, dissolve, melt, fade in, fade out, "
-                    "wobbly, weird, low quality, plastic, stock footage, video camera, boring",
-                    tooltip="Negative prompt text",
-                ),
-                IO.Int.Input(
-                    "seed",
-                    default=9,
-                    min=0,
-                    max=4294967295,
-                    step=1,
-                    display_mode=IO.NumberDisplay.number,
-                    tooltip="Random seed value",
-                    control_after_generate=False,
-                ),
-                IO.Video.Input(
-                    "video",
-                    tooltip="The reference video used to generate the output video. Must be at least 5 seconds long. "
-                    "Videos longer than 5s will be automatically trimmed. Only MP4 format supported.",
-                ),
-                IO.Combo.Input(
-                    "control_type",
-                    options=["Motion Transfer", "Pose Transfer"],
-                    default="Motion Transfer",
-                    optional=True,
-                ),
-                IO.Int.Input(
-                    "motion_intensity",
-                    default=100,
-                    min=0,
-                    max=100,
-                    step=1,
-                    tooltip="Only used if control_type is 'Motion Transfer'",
-                    optional=True,
-                ),
-                IO.Int.Input(
-                    "steps",
-                    default=60,
-                    min=60,  # steps should be greater or equal to cooldown_steps(36) + warmup_steps(24)
-                    max=100,
-                    step=1,
-                    display_mode=IO.NumberDisplay.number,
-                    tooltip="Number of inference steps",
-                ),
-            ],
-            outputs=[IO.Video.Output()],
-            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(
-                depends_on=IO.PriceBadgeDepends(),
-                expr="""{"type":"usd","usd": 2.25}""",
-            ),
-        )
-
-    @classmethod
-    async def execute(
-        cls,
-        prompt: str,
-        negative_prompt: str,
-        seed: int,
-        video: Input.Video | None = None,
-        control_type: str = "Motion Transfer",
-        motion_intensity: int | None = 100,
-        steps=60,
-        prompt_adherence=4.5,
-    ) -> IO.NodeOutput:
-        validated_video = validate_video_to_video_input(video)
-        video_url = await upload_video_to_comfyapi(cls, validated_video)
-        validate_string(prompt, min_length=1, max_length=MOONVALLEY_MAREY_MAX_PROMPT_LENGTH)
-        validate_string(negative_prompt, field_name="negative_prompt", max_length=MOONVALLEY_MAREY_MAX_PROMPT_LENGTH)
-
-        # Only include motion_intensity for Motion Transfer
-        control_params = {}
-        if control_type == "Motion Transfer" and motion_intensity is not None:
-            control_params["motion_intensity"] = motion_intensity
-
-        inference_params = MoonvalleyVideoToVideoInferenceParams(
-            negative_prompt=negative_prompt,
-            seed=seed,
-            control_params=control_params,
-            steps=steps,
-            guidance_scale=prompt_adherence,
-        )
-
-        task_creation_response = await sync_op(
-            cls,
-            endpoint=ApiEndpoint(path=API_VIDEO2VIDEO_ENDPOINT, method="POST"),
-            response_model=MoonvalleyPromptResponse,
-            data=MoonvalleyVideoToVideoRequest(
-                control_type=parse_control_parameter(control_type),
-                video_url=video_url,
-                prompt_text=prompt,
-                inference_params=inference_params,
-            ),
-        )
-        validate_task_creation_response(task_creation_response)
-        final_response = await get_response(cls, task_creation_response.id)
-        return IO.NodeOutput(await download_url_to_video_output(final_response.output_url))
-
-
-class MoonvalleyTxt2VideoNode(IO.ComfyNode):
-
-    @classmethod
-    def define_schema(cls) -> IO.Schema:
-        return IO.Schema(
-            node_id="MoonvalleyTxt2VideoNode",
-            display_name="Moonvalley Marey Text to Video",
-            category="api node/video/Moonvalley Marey",
-            description="",
-            inputs=[
-                IO.String.Input(
-                    "prompt",
-                    multiline=True,
-                ),
-                IO.String.Input(
-                    "negative_prompt",
-                    multiline=True,
-                    default="<synthetic> <scene cut> gopro, bright, contrast, static, overexposed, vignette, "
-                    "artifacts, still, noise, texture, scanlines, videogame, 360 camera, VR, transition, "
-                    "flare, saturation, distorted, warped, wide angle, saturated, vibrant, glowing, "
-                    "cross dissolve, cheesy, ugly hands, mutated hands, mutant, disfigured, extra fingers, "
-                    "blown out, horrible, blurry, worst quality, bad, dissolve, melt, fade in, fade out, "
-                    "wobbly, weird, low quality, plastic, stock footage, video camera, boring",
-                    tooltip="Negative prompt text",
-                ),
-                IO.Combo.Input(
-                    "resolution",
-                    options=[
-                        "16:9 (1920 x 1080)",
-                        "9:16 (1080 x 1920)",
-                        "1:1 (1152 x 1152)",
-                        "4:3 (1536 x 1152)",
-                        "3:4 (1152 x 1536)",
-                        "21:9 (2560 x 1080)",
-                    ],
-                    default="16:9 (1920 x 1080)",
-                    tooltip="Resolution of the output video",
-                ),
-                IO.Float.Input(
-                    "prompt_adherence",
-                    default=4.0,
-                    min=1.0,
-                    max=20.0,
-                    step=1.0,
-                    tooltip="Guidance scale for generation control",
-                ),
-                IO.Int.Input(
-                    "seed",
-                    default=9,
-                    min=0,
-                    max=4294967295,
-                    step=1,
-                    display_mode=IO.NumberDisplay.number,
-                    control_after_generate=True,
-                    tooltip="Random seed value",
-                ),
-                IO.Int.Input(
-                    "steps",
-                    default=80,
-                    min=75,  # steps should be greater or equal to cooldown_steps(75) + warmup_steps(0)
-                    max=100,
-                    step=1,
-                    tooltip="Inference steps",
-                ),
-            ],
-            outputs=[IO.Video.Output()],
-            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(
-                depends_on=IO.PriceBadgeDepends(),
-                expr="""{"type":"usd","usd": 1.5}""",
-            ),
-        )
-
-    @classmethod
-    async def execute(
-        cls,
-        prompt: str,
-        negative_prompt: str,
-        resolution: str,
-        prompt_adherence: float,
-        seed: int,
-        steps: int,
-    ) -> IO.NodeOutput:
-        validate_string(prompt, min_length=1, max_length=MOONVALLEY_MAREY_MAX_PROMPT_LENGTH)
-        validate_string(negative_prompt, field_name="negative_prompt", max_length=MOONVALLEY_MAREY_MAX_PROMPT_LENGTH)
-        width_height = parse_width_height_from_res(resolution)
-
-        inference_params = MoonvalleyTextToVideoInferenceParams(
-            negative_prompt=negative_prompt,
-            steps=steps,
-            seed=seed,
-            guidance_scale=prompt_adherence,
-            num_frames=128,
-            width=width_height["width"],
-            height=width_height["height"],
-        )
-
-        task_creation_response = await sync_op(
-            cls,
-            endpoint=ApiEndpoint(path=API_TXT2VIDEO_ENDPOINT, method="POST"),
-            response_model=MoonvalleyPromptResponse,
-            data=MoonvalleyTextToVideoRequest(prompt_text=prompt, inference_params=inference_params),
-        )
-        validate_task_creation_response(task_creation_response)
-        final_response = await get_response(cls, task_creation_response.id)
-        return IO.NodeOutput(await download_url_to_video_output(final_response.output_url))
-
-
-class MoonvalleyExtension(ComfyExtension):
-    @override
-    async def get_node_list(self) -> list[type[IO.ComfyNode]]:
-        return [
-            MoonvalleyImg2VideoNode,
-            MoonvalleyTxt2VideoNode,
-            MoonvalleyVideo2VideoNode,
-        ]
-
-
-async def comfy_entrypoint() -> MoonvalleyExtension:
-    return MoonvalleyExtension()

comfy_api_nodes/nodes_openai.py

@@ -454,7 +454,6 @@ class OpenAIGPTImage1(IO.ComfyNode):
                     step=16,
                     tooltip="Used only when `size` is 'Custom'. Must be a multiple of 16 (GPT Image 2 only).",
                     optional=True,
-                    advanced=True,
                 ),
                 IO.Int.Input(
                     "custom_height",
@@ -464,7 +463,6 @@ class OpenAIGPTImage1(IO.ComfyNode):
                     step=16,
                     tooltip="Used only when `size` is 'Custom'. Must be a multiple of 16 (GPT Image 2 only).",
                     optional=True,
-                    advanced=True,
                 ),
             ],
             outputs=[

comfy_api_nodes/nodes_topaz.py

@@ -453,7 +453,6 @@ class TopazVideoEnhance(IO.ComfyNode):
             progress_extractor=lambda x: getattr(x, "progress", 0),
             price_extractor=lambda x: (x.estimates.cost[0] * 0.08 if x.estimates and x.estimates.cost[0] else None),
             poll_interval=10.0,
-            max_poll_attempts=320,
         )
         return IO.NodeOutput(await download_url_to_video_output(final_response.download.url))
 

comfy_api_nodes/nodes_vidu.py

@@ -38,7 +38,7 @@ async def execute_task(
     cls: type[IO.ComfyNode],
     vidu_endpoint: str,
     payload: TaskCreationRequest | TaskExtendCreationRequest | TaskMultiFrameCreationRequest,
-    max_poll_attempts: int = 320,
+    max_poll_attempts: int = 480,
 ) -> list[TaskResult]:
     task_creation_response = await sync_op(
         cls,
@@ -1097,7 +1097,6 @@ class ViduExtendVideoNode(IO.ComfyNode):
                 video_url=await upload_video_to_comfyapi(cls, video, wait_label="Uploading video"),
                 images=[image_url] if image_url else None,
             ),
-            max_poll_attempts=480,
         )
         return IO.NodeOutput(await download_url_to_video_output(results[0].url))
 

comfy_api_nodes/nodes_wan.py

@@ -818,7 +818,6 @@ class WanReferenceVideoApi(IO.ComfyNode):
             response_model=VideoTaskStatusResponse,
             status_extractor=lambda x: x.output.task_status,
             poll_interval=6,
-            max_poll_attempts=280,
         )
         return IO.NodeOutput(await download_url_to_video_output(response.output.video_url))
 

comfy_api_nodes/nodes_wavespeed.py

@@ -84,7 +84,6 @@ class WavespeedFlashVSRNode(IO.ComfyNode):
             response_model=TaskResultResponse,
             status_extractor=lambda x: "failed" if x.data is None else x.data.status,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         if final_response.code != 200:
             raise ValueError(
@@ -156,7 +155,6 @@ class WavespeedImageUpscaleNode(IO.ComfyNode):
             response_model=TaskResultResponse,
             status_extractor=lambda x: "failed" if x.data is None else x.data.status,
             poll_interval=10.0,
-            max_poll_attempts=480,
         )
         if final_response.code != 200:
             raise ValueError(

comfy_api_nodes/util/client.py

@@ -148,7 +148,7 @@ async def poll_op(
     queued_statuses: list[str | int] | None = None,
     data: BaseModel | None = None,
     poll_interval: float = 5.0,
-    max_poll_attempts: int = 160,
+    max_poll_attempts: int = 480,
     timeout_per_poll: float = 120.0,
     max_retries_per_poll: int = 10,
     retry_delay_per_poll: float = 1.0,
@@ -254,7 +254,7 @@ async def poll_op_raw(
     queued_statuses: list[str | int] | None = None,
     data: dict[str, Any] | BaseModel | None = None,
     poll_interval: float = 5.0,
-    max_poll_attempts: int = 160,
+    max_poll_attempts: int = 480,
     timeout_per_poll: float = 120.0,
     max_retries_per_poll: int = 10,
     retry_delay_per_poll: float = 1.0,

comfy_extras/nodes_sdpose.py

@@ -459,27 +459,23 @@ class SDPoseKeypointExtractor(io.ComfyNode):
         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
+        model_w = int(head.heatmap_size[0]) * 4   # 192 * 4 = 768
+        model_h = int(head.heatmap_size[1]) * 4   # 256 * 4 = 1024
 
         def _resize_to_model(imgs):
-            """Aspect-preserving resize + zero-pad BHWC images to (model_h, model_w). Returns (resized_bhwc, scale, pad_top, pad_left)."""
+            """Stretch BHWC images to (model_h, model_w), model expects no aspect preservation."""
             h, w = imgs.shape[-3], imgs.shape[-2]
-            scale = min(model_h / h, model_w / w)
-            sh, sw = int(round(h * scale)), int(round(w * scale))
-            pt, pl = (model_h - sh) // 2, (model_w - sw) // 2
+            method = "area" if (model_h <= h and model_w <= w) else "bilinear"
             chw = imgs.permute(0, 3, 1, 2).float()
-            scaled = comfy.utils.common_upscale(chw, sw, sh, upscale_method="bilinear", crop="disabled")
-            padded = torch.zeros(scaled.shape[0], scaled.shape[1], model_h, model_w, dtype=scaled.dtype, device=scaled.device)
-            padded[:, :, pt:pt + sh, pl:pl + sw] = scaled
-            return padded.permute(0, 2, 3, 1), scale, pt, pl
+            scaled = comfy.utils.common_upscale(chw, model_w, model_h, upscale_method=method, crop="disabled")
+            return scaled.permute(0, 2, 3, 1), model_w / w, model_h / h
 
-        def _remap_keypoints(kp, scale, pad_top, pad_left, offset_x=0, offset_y=0):
+        def _remap_keypoints(kp, scale_x, scale_y, offset_x=0, offset_y=0):
             """Remap keypoints from model space back to original image space."""
             kp = kp.copy() if isinstance(kp, np.ndarray) else np.array(kp, dtype=np.float32)
             invalid = kp[..., 0] < 0
-            kp[..., 0] = (kp[..., 0] - pad_left) / scale + offset_x
-            kp[..., 1] = (kp[..., 1] - pad_top)  / scale + offset_y
+            kp[..., 0] = kp[..., 0] / scale_x + offset_x
+            kp[..., 1] = kp[..., 1] / scale_y + offset_y
             kp[invalid] = -1
             return kp
 
@@ -529,18 +525,18 @@ class SDPoseKeypointExtractor(io.ComfyNode):
                             continue
 
                         crop = img[:, y1:y2, x1:x2, :]  # (1, crop_h, crop_w, C)
-                        crop_resized, scale, pad_top, pad_left = _resize_to_model(crop)
+                        crop_resized, sx, sy = _resize_to_model(crop)
 
                         latent_crop = vae.encode(crop_resized)
                         kp_batch, sc_batch = _run_on_latent(latent_crop)
-                        kp = _remap_keypoints(kp_batch[0], scale, pad_top, pad_left, x1, y1)
+                        kp = _remap_keypoints(kp_batch[0], sx, sy, x1, y1)
                         img_keypoints.append(kp)
                         img_scores.append(sc_batch[0])
                 else:
-                    img_resized, scale, pad_top, pad_left = _resize_to_model(img)
+                    img_resized, sx, sy = _resize_to_model(img)
                     latent_img = vae.encode(img_resized)
                     kp_batch, sc_batch = _run_on_latent(latent_img)
-                    img_keypoints.append(_remap_keypoints(kp_batch[0], scale, pad_top, pad_left))
+                    img_keypoints.append(_remap_keypoints(kp_batch[0], sx, sy))
                     img_scores.append(sc_batch[0])
 
                 all_keypoints.append(img_keypoints)
@@ -549,12 +545,12 @@ class SDPoseKeypointExtractor(io.ComfyNode):
 
         else: # full-image mode, batched
             for batch_start in tqdm(range(0, total_images, batch_size), desc="Extracting keypoints"):
-                batch_resized, scale, pad_top, pad_left = _resize_to_model(image[batch_start:batch_start + batch_size])
+                batch_resized, sx, sy = _resize_to_model(image[batch_start:batch_start + batch_size])
                 latent_batch = vae.encode(batch_resized)
                 kp_batch, sc_batch = _run_on_latent(latent_batch)
 
                 for kp, sc in zip(kp_batch, sc_batch):
-                    all_keypoints.append([_remap_keypoints(kp, scale, pad_top, pad_left)])
+                    all_keypoints.append([_remap_keypoints(kp, sx, sy)])
                     all_scores.append([sc])
 
                 pbar.update(len(kp_batch))
@@ -727,13 +723,13 @@ class CropByBBoxes(io.ComfyNode):
                 scale = min(output_width / crop_w, output_height / crop_h)
                 scaled_w = int(round(crop_w * scale))
                 scaled_h = int(round(crop_h * scale))
-                scaled = comfy.utils.common_upscale(crop_chw, scaled_w, scaled_h, upscale_method="bilinear", crop="disabled")
+                scaled = comfy.utils.common_upscale(crop_chw, scaled_w, scaled_h, upscale_method="area", crop="disabled")
                 pad_left = (output_width  - scaled_w) // 2
                 pad_top  = (output_height - scaled_h) // 2
                 resized = torch.zeros(1, num_ch, output_height, output_width, dtype=image.dtype, device=image.device)
                 resized[:, :, pad_top:pad_top + scaled_h, pad_left:pad_left + scaled_w] = scaled
             else:  # "stretch"
-                resized = comfy.utils.common_upscale(crop_chw, output_width, output_height, upscale_method="bilinear", crop="disabled")
+                resized = comfy.utils.common_upscale(crop_chw, output_width, output_height, upscale_method="area", crop="disabled")
             crops.append(resized)
 
         if not crops:

extra_model_paths.yaml.example

@@ -1,8 +1,16 @@
 #Rename this to extra_model_paths.yaml and ComfyUI will load it
+#
+# DEPRECATED: extra_model_paths.yaml is superseded by extra_paths.yaml, which supports
+# all path configuration (system dirs, custom_nodes, and models) in a cleaner format.
+# See extra_paths.yaml.example. This file continues to work for backward compatibility.
 
 #config for comfyui
 #your base path should be either an existing comfy install or a central folder where you store all of your models, loras, etc.
 
+# When base_path is set, any standard subdirectory that exists on disk is automatically
+# registered — the explicit paths below are optional and only needed to override a
+# category or point it to a non-standard location.
+
 #comfyui:
 #     base_path: path/to/comfyui/
 #     # You can use is_default to mark that these folders should be listed first, and used as the default dirs for eg downloads

extra_paths.yaml.example

@@ -0,0 +1,84 @@
+#Rename this to extra_paths.yaml and ComfyUI will load it
+#This is the successor to extra_model_paths.yaml and supports all path configuration in one file.
+
+#config for comfyui
+#Set base_path to your ComfyUI install root. System directories (output, input, temp, user)
+#and custom_nodes are resolved relative to base_path.
+#
+#Model paths go under the 'models' block. If you only set models/base_path, all standard
+#subdirectories that exist on disk are automatically registered — no need to list them.
+#Explicit paths under models/ are optional and only needed to override a specific category
+#or point it to a non-standard location.
+
+#comfyui:
+#    base_path: path/to/comfyui/
+#    # System directories (relative to base_path, or absolute)
+#    output: output/
+#    input: input/
+#    temp: temp/
+#    user: user/
+#    # Custom nodes directory (not auto-scanned; explicit only)
+#    custom_nodes: custom_nodes/
+#    models:
+#        base_path: models/
+#        # You can use is_default to mark that these folders should be listed first,
+#        # and used as the default dirs for eg downloads
+#        #is_default: true
+#        checkpoints: checkpoints/
+#        text_encoders: |
+#             text_encoders/
+#             clip/  # legacy location still supported
+#        clip_vision: clip_vision/
+#        configs: configs/
+#        controlnet: |
+#             controlnet/
+#             t2i_adapter/
+#        diffusion_models: |
+#             diffusion_models/
+#             unet/
+#        diffusers: diffusers/
+#        embeddings: embeddings/
+#        frame_interpolation: frame_interpolation/
+#        gligen: gligen/
+#        hypernetworks: hypernetworks/
+#        latent_upscale_models: latent_upscale_models/
+#        loras: loras/
+#        model_patches: model_patches/
+#        photomaker: photomaker/
+#        style_models: style_models/
+#        upscale_models: upscale_models/
+#        vae: vae/
+#        vae_approx: vae_approx/
+#        audio_encoders: audio_encoders/
+#        classifiers: classifiers/
+
+
+#config for a1111 ui
+#all you have to do is uncomment this (remove the #) and change the base_path to where yours is installed
+
+#a111:
+#    models:
+#        base_path: path/to/stable-diffusion-webui/
+#        checkpoints: models/Stable-diffusion
+#        configs: models/Stable-diffusion
+#        vae: models/VAE
+#        loras: |
+#             models/Lora
+#             models/LyCORIS
+#        upscale_models: |
+#                      models/ESRGAN
+#                      models/RealESRGAN
+#                      models/SwinIR
+#        embeddings: embeddings
+#        hypernetworks: models/hypernetworks
+#        controlnet: models/ControlNet
+
+
+# For a full list of supported model category keys (style_models, vae_approx, hypernetworks,
+# photomaker, model_patches, audio_encoders, classifiers, etc.) see folder_paths.py.
+
+#other_ui:
+#    models:
+#        base_path: path/to/ui
+#        checkpoints: models/checkpoints
+#        gligen: models/gligen

main.py

@@ -93,9 +93,21 @@ if args.enable_manager:
 
 
 def apply_custom_paths():
-    # extra model paths
-    extra_model_paths_config_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "extra_model_paths.yaml")
-    if os.path.isfile(extra_model_paths_config_path):
+    install_dir = os.path.dirname(os.path.realpath(__file__))
+
+    # extra_paths.yaml — primary config (superset of extra_model_paths.yaml)
+    extra_paths_config_path = os.path.join(install_dir, "extra_paths.yaml")
+    extra_model_paths_config_path = os.path.join(install_dir, "extra_model_paths.yaml")
+
+    if os.path.isfile(extra_paths_config_path):
+        utils.extra_config.load_extra_path_config(extra_paths_config_path, allow_system_dirs=True)
+        if os.path.isfile(extra_model_paths_config_path):
+            logging.warning(
+                "Both extra_paths.yaml and extra_model_paths.yaml found; "
+                "ignoring the deprecated extra_model_paths.yaml. "
+                "Please remove or migrate its entries to extra_paths.yaml."
+            )
+    elif os.path.isfile(extra_model_paths_config_path):
         utils.extra_config.load_extra_path_config(extra_model_paths_config_path)
 
     if args.extra_model_paths_config:

requirements.txt

@@ -1,5 +1,5 @@
 comfyui-frontend-package==1.42.15
-comfyui-workflow-templates==0.9.65
+comfyui-workflow-templates==0.9.66
 comfyui-embedded-docs==0.4.4
 torch
 torchsde

tests-unit/utils/extra_config_test.py

@@ -301,3 +301,320 @@ def test_load_extra_path_config_no_base_path(
     actual_diffusion = folder_paths.folder_names_and_paths["diffusion_models"][0]
     assert len(actual_diffusion) == 1, "Should have one path for 'diffusion_models'."
     assert actual_diffusion[0] == os.path.abspath(expected_unet)
+
+
+@patch("yaml.safe_load")
+def test_load_extra_path_config_implicit_subdirs(
+    mock_yaml_load, clear_folder_paths, tmp_path
+):
+    """
+    When base_path is set and no explicit sub-paths are declared, any subdir
+    whose name matches a known category is auto-registered.
+    """
+    # Create real subdirs that match known categories
+    (tmp_path / "checkpoints").mkdir()
+    (tmp_path / "loras").mkdir()
+    (tmp_path / "unknown_dir").mkdir()  # not a registered category — should be ignored
+
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+        }
+    }
+    mock_yaml_load.return_value = config_data
+
+    # Pre-populate only the categories we're testing so clear_folder_paths doesn't hide them
+    folder_paths.folder_names_and_paths["checkpoints"] = ([], set())
+    folder_paths.folder_names_and_paths["loras"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_model_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")  # content ignored; yaml.safe_load is mocked
+
+    load_extra_path_config(yaml_path)
+
+    assert str(tmp_path / "checkpoints") in folder_paths.folder_names_and_paths["checkpoints"][0]
+    assert str(tmp_path / "loras") in folder_paths.folder_names_and_paths["loras"][0]
+    assert "unknown_dir" not in folder_paths.folder_names_and_paths
+
+
+@patch("yaml.safe_load")
+def test_implicit_scan_excludes_custom_nodes(
+    mock_yaml_load, clear_folder_paths, tmp_path
+):
+    """custom_nodes must never be auto-registered by the implicit scan."""
+    (tmp_path / "custom_nodes").mkdir()
+    (tmp_path / "checkpoints").mkdir()
+
+    config_data = {"comfyui": {"base_path": str(tmp_path)}}
+    mock_yaml_load.return_value = config_data
+
+    folder_paths.folder_names_and_paths["checkpoints"] = ([], set())
+    folder_paths.folder_names_and_paths["custom_nodes"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    assert str(tmp_path / "checkpoints") in folder_paths.folder_names_and_paths["checkpoints"][0]
+    assert str(tmp_path / "custom_nodes") not in folder_paths.folder_names_and_paths["custom_nodes"][0], \
+        "custom_nodes must not be auto-registered by the implicit scan"
+
+
+@patch("yaml.safe_load")
+def test_load_extra_path_config_explicit_overrides_implicit(
+    mock_yaml_load, clear_folder_paths, tmp_path
+):
+    """
+    Explicit sub-path declarations take precedence; the implicit scan must not
+    double-register a category that was already declared explicitly.
+    """
+    (tmp_path / "loras").mkdir()
+    custom_loras = tmp_path / "my_custom_loras"
+    custom_loras.mkdir()
+
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "loras": "my_custom_loras",  # explicit override
+        }
+    }
+    mock_yaml_load.return_value = config_data
+    folder_paths.folder_names_and_paths["loras"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_model_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    registered = folder_paths.folder_names_and_paths["loras"][0]
+    assert str(custom_loras) in registered
+    assert str(tmp_path / "loras") not in registered, "Implicit path must not override explicit"
+
+
+@pytest.fixture
+def save_restore_system_dirs():
+    """Save and restore folder_paths system directories around a test."""
+    saved = {
+        "output": folder_paths.get_output_directory(),
+        "input": folder_paths.get_input_directory(),
+        "temp": folder_paths.get_temp_directory(),
+        "user": folder_paths.get_user_directory(),
+    }
+    yield
+    folder_paths.set_output_directory(saved["output"])
+    folder_paths.set_input_directory(saved["input"])
+    folder_paths.set_temp_directory(saved["temp"])
+    folder_paths.set_user_directory(saved["user"])
+
+
+@patch("yaml.safe_load")
+def test_system_dir_keys(mock_yaml_load, save_restore_system_dirs, tmp_path):
+    """System directory keys (output, input, temp, user) call set_*_directory()."""
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "output": "my_output/",
+            "input": "my_input/",
+            "temp": "my_temp/",
+            "user": "my_user/",
+        }
+    }
+    mock_yaml_load.return_value = config_data
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path, allow_system_dirs=True)
+
+    assert folder_paths.get_output_directory() == os.path.normpath(str(tmp_path / "my_output"))
+    assert folder_paths.get_input_directory() == os.path.normpath(str(tmp_path / "my_input"))
+    assert folder_paths.get_temp_directory() == os.path.normpath(str(tmp_path / "my_temp"))
+    assert folder_paths.get_user_directory() == os.path.normpath(str(tmp_path / "my_user"))
+
+
+@patch("yaml.safe_load")
+def test_system_dir_keys_not_applied_for_legacy(mock_yaml_load, save_restore_system_dirs, tmp_path):
+    """System directory keys are ignored when allow_system_dirs=False (legacy extra_model_paths.yaml)."""
+    original_output = folder_paths.get_output_directory()
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "output": "my_output/",
+        }
+    }
+    mock_yaml_load.return_value = config_data
+
+    yaml_path = str(tmp_path / "extra_model_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)  # allow_system_dirs defaults to False
+
+    # output should be unchanged — treated as a model category, not a system dir
+    assert folder_paths.get_output_directory() == original_output
+
+
+@patch("yaml.safe_load")
+def test_nested_models_block(mock_yaml_load, clear_folder_paths, tmp_path):
+    """Nested models: block registers model paths relative to models/base_path."""
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "models": {
+                "base_path": "models/",
+                "checkpoints": "checkpoints/",
+                "loras": "loras/",
+            },
+        }
+    }
+    mock_yaml_load.return_value = config_data
+
+    folder_paths.folder_names_and_paths["checkpoints"] = ([], set())
+    folder_paths.folder_names_and_paths["loras"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    expected_ckpt = os.path.normpath(str(tmp_path / "models" / "checkpoints"))
+    expected_loras = os.path.normpath(str(tmp_path / "models" / "loras"))
+    assert expected_ckpt in folder_paths.folder_names_and_paths["checkpoints"][0]
+    assert expected_loras in folder_paths.folder_names_and_paths["loras"][0]
+
+
+@patch("yaml.safe_load")
+def test_nested_models_is_default(mock_yaml_load, clear_folder_paths, tmp_path):
+    """is_default under models: applies to all model paths in that block."""
+    config_data = {
+        "comfyui": {
+            "models": {
+                "base_path": str(tmp_path),
+                "is_default": True,
+                "checkpoints": "checkpoints/",
+            },
+        }
+    }
+    mock_yaml_load.return_value = config_data
+    folder_paths.folder_names_and_paths["checkpoints"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    mock_add = Mock()
+    with patch.object(folder_paths, "add_model_folder_path", mock_add):
+        load_extra_path_config(yaml_path)
+
+    call = mock_add.call_args_list[0]
+    assert call.args[0] == "checkpoints"
+    assert call.args[2] is True, "is_default under models: must be passed as True"
+
+
+@patch("yaml.safe_load")
+def test_nested_models_multipath(mock_yaml_load, clear_folder_paths, tmp_path):
+    """Multi-line path values inside models: register multiple paths per category."""
+    config_data = {
+        "comfyui": {
+            "models": {
+                "base_path": str(tmp_path),
+                "text_encoders": "text_encoders/\nclip/",
+            },
+        }
+    }
+    mock_yaml_load.return_value = config_data
+    folder_paths.folder_names_and_paths["text_encoders"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    registered = folder_paths.folder_names_and_paths["text_encoders"][0]
+    assert os.path.normpath(str(tmp_path / "text_encoders")) in registered
+    assert os.path.normpath(str(tmp_path / "clip")) in registered
+
+
+@patch("yaml.safe_load")
+def test_nested_models_auto_scan(mock_yaml_load, clear_folder_paths, tmp_path):
+    """models: with only base_path auto-scans for known categories that exist on disk."""
+    (tmp_path / "models" / "checkpoints").mkdir(parents=True)
+    (tmp_path / "models" / "loras").mkdir()
+
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "models": {"base_path": "models/"},
+        }
+    }
+    mock_yaml_load.return_value = config_data
+    folder_paths.folder_names_and_paths["checkpoints"] = ([], set())
+    folder_paths.folder_names_and_paths["loras"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    assert os.path.normpath(str(tmp_path / "models" / "checkpoints")) in \
+        folder_paths.folder_names_and_paths["checkpoints"][0]
+    assert os.path.normpath(str(tmp_path / "models" / "loras")) in \
+        folder_paths.folder_names_and_paths["loras"][0]
+
+
+@patch("yaml.safe_load")
+def test_explicit_custom_nodes_key(mock_yaml_load, clear_folder_paths, tmp_path):
+    """Explicit custom_nodes key in a block registers the path via add_model_folder_path."""
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "custom_nodes": "my_nodes/",
+        }
+    }
+    mock_yaml_load.return_value = config_data
+    folder_paths.folder_names_and_paths["custom_nodes"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    assert os.path.normpath(str(tmp_path / "my_nodes")) in \
+        folder_paths.folder_names_and_paths["custom_nodes"][0]
+
+
+@patch("yaml.safe_load")
+def test_nested_models_inherits_block_base(mock_yaml_load, clear_folder_paths, tmp_path):
+    """models: block without its own base_path inherits the outer block's base_path."""
+    config_data = {
+        "comfyui": {
+            "base_path": str(tmp_path),
+            "is_default": True,
+            "models": {
+                "checkpoints": "models/checkpoints/",
+            },
+        }
+    }
+    mock_yaml_load.return_value = config_data
+    folder_paths.folder_names_and_paths["checkpoints"] = ([], set())
+
+    yaml_path = str(tmp_path / "extra_paths.yaml")
+    with open(yaml_path, "w") as f:
+        f.write("")
+
+    load_extra_path_config(yaml_path)
+
+    expected = os.path.normpath(str(tmp_path / "models" / "checkpoints"))
+    paths = folder_paths.folder_names_and_paths["checkpoints"][0]
+    assert expected in paths
+    # is_default inherited: path should be at index 0
+    assert paths[0] == expected

utils/extra_config.py

@@ -1,34 +1,101 @@
+from __future__ import annotations
+
 import os
 import yaml
 import folder_paths
 import logging
 
-def load_extra_path_config(yaml_path):
+_SYSTEM_DIR_KEYS = frozenset({"output", "input", "temp", "user"})
+
+
+def _resolve_base(raw: str, parent_base: str | None, yaml_dir: str) -> str:
+    """Resolve a base_path value: expand vars/user, join onto parent_base or yaml_dir if relative."""
+    raw = os.path.expandvars(os.path.expanduser(raw))
+    if not os.path.isabs(raw):
+        anchor = parent_base if parent_base else yaml_dir
+        raw = os.path.abspath(os.path.join(anchor, raw))
+    return os.path.normpath(raw)
+
+
+def _add_model_paths(category: str, raw_value: str, base: str | None, yaml_dir: str, is_default: bool) -> None:
+    """Split a (possibly multi-line) path value and register each path as a model folder."""
+    for raw in str(raw_value).split("\n"):
+        raw = raw.strip()
+        if not raw:
+            continue
+        if base and not os.path.isabs(raw):
+            full_path = os.path.join(base, raw)
+        elif not os.path.isabs(raw):
+            full_path = os.path.abspath(os.path.join(yaml_dir, raw))
+        else:
+            full_path = raw
+        normalized = os.path.normpath(full_path)
+        logging.info("Adding extra search path %s %s", category, normalized)
+        folder_paths.add_model_folder_path(category, normalized, is_default)
+
+
+def _implicit_scan(base: str, exclude: set[str], is_default: bool) -> None:
+    """Auto-register base/<category>/ for known model categories that exist on disk.
+
+    custom_nodes and system directory keys are always excluded from the scan.
+    """
+    skip = _SYSTEM_DIR_KEYS | {"custom_nodes"} | exclude
+    for category in folder_paths.folder_names_and_paths:
+        if category in skip:
+            continue
+        path = os.path.normpath(os.path.join(base, category))
+        if os.path.isdir(path):
+            logging.info("Adding extra search path %s %s", category, path)
+            folder_paths.add_model_folder_path(category, path, is_default)
+
+
+def load_extra_path_config(yaml_path: str, allow_system_dirs: bool = False) -> None:
     with open(yaml_path, 'r', encoding='utf-8') as stream:
         config = yaml.safe_load(stream)
     yaml_dir = os.path.dirname(os.path.abspath(yaml_path))
-    for c in config:
-        conf = config[c]
+
+    for _block_name, conf in config.items():
         if conf is None:
             continue
-        base_path = None
+
+        # Pop block-level meta keys (preserved for flat backward-compat style)
+        block_base = None
         if "base_path" in conf:
-            base_path = conf.pop("base_path")
-            base_path = os.path.expandvars(os.path.expanduser(base_path))
-            if not os.path.isabs(base_path):
-                base_path = os.path.abspath(os.path.join(yaml_dir, base_path))
-        is_default = False
-        if "is_default" in conf:
-            is_default = conf.pop("is_default")
-        for x in conf:
-            for y in conf[x].split("\n"):
-                if len(y) == 0:
-                    continue
-                full_path = y
-                if base_path:
-                    full_path = os.path.join(base_path, full_path)
-                elif not os.path.isabs(full_path):
-                    full_path = os.path.abspath(os.path.join(yaml_dir, y))
-                normalized_path = os.path.normpath(full_path)
-                logging.info("Adding extra search path {} {}".format(x, normalized_path))
-                folder_paths.add_model_folder_path(x, normalized_path, is_default)
+            block_base = _resolve_base(conf.pop("base_path"), None, yaml_dir)
+        block_is_default = bool(conf.pop("is_default", False))
+
+        has_models_block = False
+        flat_model_keys: set[str] = set()
+
+        for key, value in conf.items():
+            if allow_system_dirs and key in _SYSTEM_DIR_KEYS:
+                # System directory override → set_*_directory()
+                path = _resolve_base(str(value).strip(), block_base, yaml_dir)
+                logging.info("Setting %s directory to %s", key, path)
+                getattr(folder_paths, f"set_{key}_directory")(path)
+
+            elif key == "custom_nodes":
+                _add_model_paths("custom_nodes", value, block_base, yaml_dir, block_is_default)
+
+            elif key == "models" and isinstance(value, dict):
+                # New nested style: models: { base_path, is_default, <categories> }
+                has_models_block = True
+                models_conf = dict(value)
+                models_base = block_base
+                if "base_path" in models_conf:
+                    models_base = _resolve_base(models_conf.pop("base_path"), block_base, yaml_dir)
+                models_is_default = bool(models_conf.pop("is_default", block_is_default))
+                explicit: set[str] = set(models_conf.keys())
+                for cat, raw in models_conf.items():
+                    _add_model_paths(cat, raw, models_base, yaml_dir, models_is_default)
+                if models_base:
+                    _implicit_scan(models_base, explicit, models_is_default)
+
+            else:
+                # Flat model key — backward-compat style
+                _add_model_paths(key, value, block_base, yaml_dir, block_is_default)
+                flat_model_keys.add(key)
+
+        # Flat-style implicit scan (only when no nested models: block)
+        if block_base and not has_models_block:
+            _implicit_scan(block_base, flat_model_keys, block_is_default)