Comfy-Org/ComfyUI · error · ValueError

Output size ({output_width}x{output_height} = {output_pixels

Error message

Output size ({output_width}x{output_height} = {output_pixels:,} pixels) exceeds maximum allowed size of {MAX_PIXELS_GENERATIVE:,} pixels ({MAX_MP_GENERATIVE}MP). Enable auto_downscale or use a smaller input image or a lower upscale factor.

What it means

Thrown by Hitpaw photo enhancement when width*height*scale^2 exceeds MAX_PIXELS_GENERATIVE and auto_downscale is disabled. With auto_downscale=True the node instead walks candidate scales [4,2,1] and optionally downscales the input by up to 2x to fit; the error is the explicit refusal path when the user wants no automatic shrinking. The message states the computed output size and the exact pixel cap.

Source

Thrown at comfy_api_nodes/nodes_hitpaw.py:153

                    scale_output_pixels = input_pixels * candidate * candidate
                    if scale_output_pixels <= MAX_PIXELS_GENERATIVE:
                        scale = candidate
                        max_input_pixels = None
                        break
                    # Check if we can downscale input by at most 2x to fit
                    downscale_ratio = math.sqrt(scale_output_pixels / MAX_PIXELS_GENERATIVE)
                    if downscale_ratio <= 2.0:
                        scale = candidate
                        max_input_pixels = MAX_PIXELS_GENERATIVE // (candidate * candidate)
                        break

                if max_input_pixels is not None:
                    image = downscale_image_tensor(image, total_pixels=max_input_pixels)
                upscale_factor = scale
            else:
                output_width = width * requested_scale
                output_height = height * requested_scale
                raise ValueError(
                    f"Output size ({output_width}x{output_height} = {output_pixels:,} pixels) "
                    f"exceeds maximum allowed size of {MAX_PIXELS_GENERATIVE:,} pixels ({MAX_MP_GENERATIVE}MP). "
                    f"Enable auto_downscale or use a smaller input image or a lower upscale factor."
                )

        initial_res = await sync_op(
            cls,
            ApiEndpoint(path="/proxy/hitpaw/api/photo-enhancer", method="POST"),
            response_model=TaskCreateResponse,
            data=ImageEnhanceTaskCreateRequest(
                model_name=f"{model}_{upscale_factor}x",
                img_url=await upload_image_to_comfyapi(cls, image, total_pixels=None),
            ),
            wait_label="Creating task",
            final_label_on_success="Task created",
        )
        if initial_res.code != 200:
            raise ValueError(f"Task creation failed with code {initial_res.code}: {initial_res.message}")

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Enable auto_downscale — the node picks the largest fitting scale (and at most a 2x input downscale) automatically
2. Lower upscale_factor to 1 or 2 so output stays under the cap
3. Pre-downscale the input image so input_pixels * scale^2 fits within MAX_PIXELS_GENERATIVE

Example fix

# before
auto_downscale = False; upscale_factor = 4  # 8000x8000 input -> error
# after
auto_downscale = True  # node auto-selects a fitting scale, e.g. 2x or 1x

from comfy_api_nodes.utils import get_image_dimensions
from comfy_api_nodes.nodes_hitpaw import MAX_PIXELS_GENERATIVE

def enhance_output_fits(image, scale: int) -> bool:
    h, w = get_image_dimensions(image)
    return h * w * scale * scale <= MAX_PIXELS_GENERATIVE

When it happens

Trigger: Large input combined with upscale_factor 2/4 such that output pixels exceed the cap; e.g. a 6000x6000 image with 4x selected. Only raised when auto_downscale=False.

Common situations: Users upscaling already-high-resolution outputs from generation pipelines, or chaining enhancers where a previous 4x pass feeds this one.

Related errors

• Selected resolution {resolution} ({target_size}x{target_size
• Output size ({width * requested_scale}x{height * requested_s
• Unsupported process_res_method: {method}
• At least 4 points are required to compute a homography.
• Unknown reference view selection strategy: {strategy!r}. Mus

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/400d7f10106abe14. Report an issue: GitHub.