Image Pad

Padding a canvas is a two-second job in any image editor, so a node that does it needs a reason to exist. UC_ImagePad's reason: it hands you a placement mask alongside the padded image. That mask is what tells an outpainting or compositing pass exactly where the new canvas is and where the original image sits - which is the difference between an outpaint that fills the right region and one that guesses.

It's the outpainting prep node. You want to extend a 1024 image to 1536 for a wider composition; you pad the sides, and the generated mask (value 1 in the padded zone, 0 over the original) becomes the inpaint/outpaint mask that says "only regenerate the new area." Without that, you'd be hand-drawing a mask that's trivially computable - and this node computes it for you, exactly aligned.

How it works

Two padding modes:

• Explicit: left/right/top/bottom add those pixels to each side (defaults 0).
• Target size: connect target_width and target_height and it proportionally fits and centers the image inside that canvas, using extra_padding as an inset margin.

pad_mode picks the fill: edge extends the border pixels (the standard look for outpaint prep), edge_pixel keeps the original edge pixel's color, and color fills with a solid - where color accepts RGB triples, 0–255 or 0.0–1.0, a color name, or a hex code.

The mask input is optional: if you connect your own mask, it gets padded with value 1 outside the image; disconnected, a mask for the added border is created for you. That last clause is the feature - the output mask socket always exists and always describes the padding.

The inputs that matter

For beginners: image, left/right (most outpaints pad sides), pad_mode = edge, and then wire the output mask into your outpaint's mask input. target_width/target_height are the alternative when you think "I want exactly 1536 wide" rather than "add 256 left and 256 right." Outputs: image and mask.

Installing it

ComfyUI Manager → search "ComfyUI-UtilsCollection", or:

cd ComfyUI/custom_nodes
git clone https://github.com/silveroxides/ComfyUI-UtilsCollection

Restart. Deps: opencv-python, typing-extensions, no models. Newer-API pack - keep ComfyUI updated.

The classic mix-up: connecting one of target_width/target_height does nothing useful - both must be connected for target mode, and the tooltip is explicit that the explicit-pad inputs are ignored when target dimensions are connected. If your padding ignores the numbers you typed, check which mode you're in.