Nuke Color Bars

NukeColorBars is the test-pattern generator in this pack, and if you've never needed a test pattern, it sounds useless. Then you need one: you're checking whether a monitor or a save pipeline is grading your images, comparing two output paths, or just building a test workflow that doesn't depend on a model. A standard color bar pattern is the reference - every image tool in the world knows what SMPTE bars are supposed to look like, which makes them the perfect control card.

This is a generator node with no image input - you set dimensions and a pattern, and out comes an IMAGE. Four patterns: smpte (the classic broadcast bars with the lower color band), rgb_bars (simple red/green/blue bars), primary_colors, and grayscale (a smooth ramp from black to white). The grayscale one is quietly the most useful for colorists - a clean ramp shows banding, gamma shifts, and clipping at a glance.

How it works

It just builds the pattern as a tensor - no model involved, deterministic, instant. Give it a width, a height, and a batch_size (if you want N identical frames, useful when you're piping into a video workflow that expects a batch). Every frame is identical, so it doubles as a reference when you're testing frame-accurate outputs.

The inputs that matter

• width / height - output resolution, 64 to 4096.
• pattern - smpte, rgb_bars, primary_colors, or grayscale.
• batch_size - how many copies of the pattern to generate, 1 to 64.

Output is a single IMAGE (a batch of identical frames if batch_size is above 1).

When you'd reach for it

• Sanity-checking a color pipeline - push SMPTE bars through your save node, reload the file, and compare. If the bars don't match, your pipeline is shifting color.
• Testing grading nodes - apply a grade to the bars and you can see exactly what the grade does to known color values, instead of guessing on a generated image.
• Building test workflows - a generator node means your test graph doesn't depend on a checkpoint or a seed.

Installing it

It's part of the Nuke Nodes pack:

cd ComfyUI/custom_nodes
git clone https://github.com/sumitchatterjee13/nuke-nodes-comfyui.git nuke-nodes
cd nuke-nodes
pip install -r requirements.txt

Or search "Nuke Nodes" in ComfyUI Manager and restart. It's under the Nuke category, pure PyTorch tensor generation - zero special dependencies.

Common issues

• The bars look different on screen than expected - that's the point of using them: your display or your save pipeline is doing something. Bars are a reference, not a decoration.
• "Why would I ever need this" - you won't, until you're debugging a color shift, and then it's the fastest reference you have. Think of it as a calibration tool you hope to need rarely.

Keep this node in the back of your toolkit for the day something about your color pipeline feels off - it'll tell you the truth fast.