Mask Overlay Comparer

Masks are the part of a workflow you're most likely to trust blindly and regret later. A soft mask from a SAM segmenter or a tagger can have faint gradients, stray specks, and coverage that doesn't match what you think you selected - and a Preview Image node shows you a grayscale blob that's genuinely hard to read against the image. IPT-MaskOverlayComparer fixes the visibility problem: it draws the mask over the actual image and gives you a slider to compare the two views in one place.

It's the mask-tooling equivalent of rgthree's Image Comparer, but purpose-built for exactly this question: "is my mask covering the right region, and does its softness make sense?" The pack's own SAM3 node deliberately outputs non-binarized soft masks, which makes a node like this almost mandatory - you cannot judge a soft mask by squinting.

How it works

Give it an image and a mask. It builds a slider-based preview where one side is the clean background image and the other is the image with the mask overlaid. The rendering rule is the part to internalize: pixels where the mask is 0.0 stay unchanged, pixels where it's 1.0 render black, and in-between values get a blue-tinted overlay. So a fully-masked area reads as a black region, partial coverage reads as blue haze, and you can see at a glance both where the mask is and how soft it is. The mask is visualized with its raw 0.0–1.0 values - no binarization, which is the point.

Both inputs pass through as outputs (image and mask), so the node can sit inline in a graph as a "checkpoint" without changing the data flow. It's an output node, so it works fine as a terminal preview too.

Inputs

• image (required, IMAGE) - the background for the preview. With a batched image, only the first item is shown in the comparer.
• mask (required, MASK) - the soft or binary mask to overlay.

For best results the image and mask should be the same resolution - the tooltip says it plainly, and mismatched sizes make the overlay misleading.

Installing it

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/kinorax/comfyui-info-prompt-toolkit.git
cd comfyui-info-prompt-toolkit
pip install -r requirements.txt

Restart, or ComfyUI Manager → "ComfyUI-Info-Prompt-Toolkit". ComfyUI 0.17.0+. No models.

Common issues

If you're feeding it a batched image and the preview shows "the wrong frame," remember it only renders the first item - you may need to index to the frame you actually care about. And if the blue overlay looks wrong, check that your mask isn't inverted relative to your expectation; the convention here is mask=1.0 renders black. That's a pipeline convention mismatch, not a bug, and half of "my mask is wrong" complaints turn out to be exactly that. Also, mask=1.0 renders black is worth remembering when you're thresholding: this node shows you the truth, and if the truth is "everything's black," you probably thresholded wrong.