Comfy-Org/ComfyUI · error · ValueError

Invalid image dimensions

Error message

Invalid image dimensions

What it means

ValueError from validate_images_aspect_ratio_closeness when either of the two compared images has a non-positive dimension (min(w1,h1,w2,h2) <= 0). The function computes closeness C = max(ar1,ar2)/min(ar1,ar2), which is undefined for degenerate tensors, so it rejects them before dividing. Unlike validate_image_aspect_ratio, the message does not include the values.

Source

Thrown at comfy_api_nodes/util/validation_utils.py:70

def validate_images_aspect_ratio_closeness(
    first_image: torch.Tensor,
    second_image: torch.Tensor,
    min_rel: float,  # e.g. 0.8
    max_rel: float,  # e.g. 1.25
    *,
    strict: bool = False,  # True -> (min, max); False -> [min, max]
) -> float:
    """
    Validates that the two images' aspect ratios are 'close'.
    The closeness factor is C = max(ar1, ar2) / min(ar1, ar2)  (C >= 1).
    We require C <= limit, where limit = max(max_rel, 1.0 / min_rel).

    Returns the computed closeness factor C.
    """
    w1, h1 = get_image_dimensions(first_image)
    w2, h2 = get_image_dimensions(second_image)
    if min(w1, h1, w2, h2) <= 0:
        raise ValueError("Invalid image dimensions")
    ar1 = w1 / h1
    ar2 = w2 / h2
    closeness = max(ar1, ar2) / min(ar1, ar2)
    limit = max(max_rel, 1.0 / min_rel)
    if (closeness >= limit) if strict else (closeness > limit):
        raise ValueError(
            f"Aspect ratios must be close: ar1/ar2={ar1/ar2:.2g}, "
            f"allowed range {min_rel}–{max_rel} (limit {limit:.2g})."
        )
    return closeness


def validate_aspect_ratio_string(
    aspect_ratio: str,
    min_ratio: tuple[float, float] | None = None,  # e.g. (1, 4)
    max_ratio: tuple[float, float] | None = None,  # e.g. (4, 1)
    *,
    strict: bool = False,  # True -> (min, max); False -> [min, max]

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Verify both tensors have strictly positive H and W before pairing them.
2. Fix the crop/slice code that emptied one image.
3. Add a shape assertion in the calling node so failures point at the right tensor.

def pair_has_positive_dims(a: torch.Tensor, b: torch.Tensor) -> bool:
    return all(min(t.shape[-2], t.shape[-3]) > 0 for t in (a, b))

When it happens

Trigger: Calling validate_images_aspect_ratio_closeness(first, second, min_rel, max_rel) where first or second has a zero-sized dimension (bad slice, empty batch element).

Common situations: Comparing a generated image with a reference that was cropped to nothing; empty tensor from a failed loader; slicing bugs in custom pairing nodes.

Related errors

• Invalid image tensor shape.
• Invalid image dimensions: {w}x{h}
• JoyImage reference inputs must contain one image each
• The maximum number of reference images is 10.
• sync.so rejects images above 4K (4096x2160); got {width}x{he

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/06c4b12ee3479be4. Report an issue: GitHub.