Comfy-Org/ComfyUI · error · ValueError

The maximum number of reference images allowed with a video

Error message

The maximum number of reference images allowed with a video input is 4.

What it means

Raised by OmniProVideoToVideoNode.execute when more than 4 reference images accompany a reference video. With a video input the Kling omni API tightens the image reference cap from 7 to 4; the check runs after video duration/dimension validation and before uploads.

Source

Thrown at comfy_api_nodes/nodes_kling.py:1217

        model_name: str,
        prompt: str,
        aspect_ratio: str,
        duration: int,
        reference_video: Input.Video,
        keep_original_sound: bool,
        reference_images: Input.Image | None = None,
        resolution: str = "1080p",
        seed: int = 0,
    ) -> IO.NodeOutput:
        _ = seed
        prompt = normalize_omni_prompt_references(prompt)
        validate_string(prompt, min_length=1, max_length=2500)
        validate_video_duration(reference_video, min_duration=3.0, max_duration=10.05)
        validate_video_dimensions(reference_video, min_width=720, min_height=720, max_width=2160, max_height=2160)
        image_list: list[OmniParamImage] = []
        if reference_images is not None:
            if get_number_of_images(reference_images) > 4:
                raise ValueError("The maximum number of reference images allowed with a video input is 4.")
            for i in reference_images:
                validate_image_dimensions(i, min_width=300, min_height=300)
                validate_image_aspect_ratio(i, (1, 2.5), (2.5, 1))
            for i in await upload_images_to_comfyapi(cls, reference_images, wait_label="Uploading reference image"):
                image_list.append(OmniParamImage(image_url=i))
        video_list = [
            OmniParamVideo(
                video_url=await upload_video_to_comfyapi(cls, reference_video, wait_label="Uploading reference video"),
                refer_type="feature",
                keep_original_sound="yes" if keep_original_sound else "no",
            )
        ]
        response = await sync_op(
            cls,
            ApiEndpoint(path="/proxy/kling/v1/videos/omni-video", method="POST"),
            response_model=TaskStatusResponse,
            data=OmniProReferences2VideoRequest(
                model_name=model_name,

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Trim reference_images to at most 4 when a reference video is connected
2. Move subject/style cues from extra images into the prompt or the reference video itself

Example fix

# before
reference_images = character_refs.images  # 6 images
reference_video = load_ref_video.output

# after
reference_images = character_refs.images[:4]  # <= 4 with video input
reference_video = load_ref_video.output

def reference_count_ok_with_video(reference_images) -> bool:
    return reference_images is None or get_number_of_images(reference_images) <= 4

When it happens

Trigger: reference_video connected (3.0-10.05s, 720-2160px) plus a reference_images batch of 5 or more on the video-to-video omni node.

Common situations: Reusing a 6-image reference set that worked on the image-to-video node on the video-driven variant, which has the lower cap.

Related errors

• The maximum number of reference images allowed is 6.
• The maximum number of reference images is 7.
• The maximum number of reference images is 10.
• The 'end_frame' input cannot be used simultaneously with 're
• The 'end_frame' input cannot be used simultaneously with sto

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/0fa243cbfb5b3eef. Report an issue: GitHub.