UR Demo Image

UR Demo Image is the pack's built-in smoke test. Instead of making you hunt down a photo to verify an install, it generates a synthetic test frame on the fly: a soft RGB gradient with a solid red block sitting in the upper-left quadrant. That block is there on purpose - it gives you something obvious to select when you're testing the UniversalRemove pipeline and you want to see the mask land.

It's the first node in the example workflow the author ships, and it exists for a boring but real reason: most "is this working?" debugging in ComfyUI is actually "did I install the thing correctly?" debugging. This node answers that in one click, with no file path, no model download, and no dependency beyond what ComfyUI core already has.

What it does

It's about as simple as a node gets. Three inputs, one output.

• width / height - the frame size. Defaults to 768×512, steps of 8 (so it plays nice with latent-friendly dimensions).
• batch_size - how many frames to generate, 1 to 256. Crank this up if you want to test how a mask behaves across a video-style frame batch rather than a single image.

The output image is a standard IMAGE batch, so it plugs straight into any node that takes images - including the rest of this pack's pipeline.

Why you'd actually reach for it

Two cases. First, install verification: drop the pack in, add UR Demo Image → a Preview node, hit run, and if you see the gradient you know the custom node registered correctly without needing a real asset. Second, as a scaffold: when you're learning the UniversalRemove flow (regions → mask → keyframe prep → composite), a synthetic flat image is the perfect ground truth. A real photo has textures and lighting that make it hard to tell whether the mask is slightly off; a flat gradient makes every misalignment scream.

Installing it

This pack has zero runtime dependencies and downloads no model weights, which is the whole point of its design. Install it with ComfyUI Manager (search "ComfyUI-UniversalRemove") or the manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/fengwk/ComfyUI-UniversalRemove

Then restart ComfyUI. That's the entire install - no pip install, no weights folder, no API key. The only extra file in the repo is a dev-only requirements file for the author's pytest suite.

Common issues

Honestly, the failure mode here is expecting too much from it. It's a flat synthetic gradient, not a photograph - it exists to prove wiring works, not to stress-test a remove model. Don't judge mask quality against it. Also worth knowing: the red block is positioned relative to width and height, so if you set a wildly non-square aspect ratio it still renders, just with the block somewhere you might not expect. If you get a black screen, that's a wiring problem upstream, not this node - check that you actually connected the output to a Preview or Save Image node.

Think of it as the pack saying "hello" - a cheap, honest signal that everything after it has a chance of working.