Comfy-Org/ComfyUI · error · ValueError

Compositor canvas {cw}x{ch} exceeds the maximum supported si

Error message

Compositor canvas {cw}x{ch} exceeds the maximum supported size of {MAX_RESOLUTION}x{MAX_RESOLUTION}

What it means

Before allocating the float32 numpy canvas (ch x cw x 4), composite_from_state checks the resolved canvas dimensions against ComfyUI's MAX_RESOLUTION constant. A canvas larger than that on either axis would imply a multi-gigabyte RGBA buffer, so it is rejected up front with a message naming the offending size and the cap.

Source

Thrown at comfy_extras/nodes_compositor.py:430

def _fill_background(canvas: np.ndarray, background: dict) -> np.ndarray:
    layer = np.empty(canvas.shape, dtype=np.float32)
    layer[..., :3] = srgb_to_linear(
        np.array(hex_to_rgb(background["color"]), dtype=np.float32) / 255.0
    )
    layer[..., 3] = 1.0
    return blend_composite(
        resolve_mode("normal"), canvas, layer, background["opacity"]
    )


def composite_from_state(
    tensors: list[torch.Tensor],
    state: dict,
    alphas: list[torch.Tensor | None],
) -> torch.Tensor:
    cw, ch = state["canvas"]
    if cw > MAX_RESOLUTION or ch > MAX_RESOLUTION:
        raise ValueError(
            f"Compositor canvas {cw}x{ch} exceeds the maximum supported size of "
            f"{MAX_RESOLUTION}x{MAX_RESOLUTION}"
        )
    canvas = np.zeros((ch, cw, 4), dtype=np.float32)
    background = state.get("background")
    if background is not None and background["visible"] and background["opacity"] > 0:
        canvas = _fill_background(canvas, background)
    layers = state["layers"]
    order = state.get("order") or range(len(tensors))
    for index in order:
        if index < 0 or index >= len(tensors):
            continue
        tensor = tensors[index]
        entry = layers[index] if index < len(layers) else None
        params = _layer_params(entry, tensor.shape[2], tensor.shape[1])
        if not params["visible"]:
            continue
        img = _prepare_layer_bitmap(

View on GitHub (pinned to 1c6d8d45b3)

Solutions

Cap canvas width/height to MAX_RESOLUTION or below in the document (usually 16384).
Fix the canvas computation upstream — look for a doubled scale factor or concatenated digits.
Tile the work: composite at a smaller canvas and upscale, or split into regions if true gigapixel output is needed.

Example fix

// before
"canvas": [32768, 32768]
// after
"canvas": [16384, 16384]

Defensive patterns

Strategy: validation

Validate before calling

from nodes import MAX_RESOLUTION
def clamp_canvas(doc):
    cw, ch = doc.get("canvas", (0, 0))
    if cw > MAX_RESOLUTION or ch > MAX_RESOLUTION:
        raise ValueError("canvas too large; fix source data")
    return doc

Prevention

Validate canvas dims against MAX_RESOLUTION (16384) before building the document.
Audit scale-factor math that computes canvas size multiplicatively.
Split gigapixel composites into tiles.

When it happens

Trigger: A LAYERS document whose canvas field (or derived canvas from the background image) resolves to width or height greater than MAX_RESOLUTION (ComfyUI's standard 16384). For example a canvas of [32768, 512] or layer dimensions propagating a huge size.

Common situations: Typos in canvas JSON (extra digit), documents authored for a different engine with huge canvases, or workflows that compute canvas size by multiplying scale factors and accidentally double the intended resolution.

Related errors

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/9e9a6c2471206c7b. Report an issue: GitHub.