PalletteGenerator(Badman)

(Yes, it's spelled "Pallette" with two Ls in the pack itself - a small, harmless typo baked into the class name, and one more sign this is a personal utility pack rather than something that's been through a lot of eyes.) What it does is exactly what it says: generates an image made of random colors, laid out as a grid, useful as a quick palette reference or as procedural test material.

Why you'd want this

There's a genuine, if unglamorous, use case for a node that just spits out random colors on demand: quick synthetic test images for checking a downstream node actually works before you burn a real generation on it, a fast source of color-reference swatches when you're eyeballing a palette, or raw material to feed into this pack's own color tools - a random palette makes a decent stress-test input for something like Color Match or LABColorTransfer when you want to see how they behave across a wide, unpredictable range of colors rather than just your one carefully-chosen test image. It's also just handy as procedural background or texture material when you want visual noise that isn't tied to any particular seed image.

Inputs and outputs

Four required fields:

• width, height (integer, min 1, default 1024) - the output image dimensions.
• batch_size (integer, 1–4096, default 1) - how many palette images to generate at once. That upper bound is large enough to suggest this is meant to support generating many palette variations in a single run, useful if you're sweeping across a range of random palettes for testing rather than eyeballing one at a time.
• num_colors (integer, min 1, default 4) - how many distinct colors make up the grid. Higher values give you a finer, busier grid of smaller color swatches; lower values give you a small handful of large, easy-to-eyeball blocks.

One output: the generated IMAGE.

Installing it

ComfyUI Manager: search ComfyUI-BadmanNodes, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/MariusKM/ComfyUI-BadmanNodes

Nothing else needed - this generates colors procedurally, no model weights, no dependencies beyond the pack itself.

Where people get tripped up

There's no seed input on this node per its schema, so if you need a reproducible random palette rather than a fresh one every run, this isn't the tool for that - it's built for quick, disposable reference material, not deterministic testing. If reproducibility matters for your use case, you'll want to generate the palette once and save it as an actual image file rather than regenerating it in the graph each time. Second, a large batch_size combined with large width/height values adds up fast in memory - 4096 batch items at a substantial resolution is a lot of image data to hold at once, so scale batch size down if you only actually need a handful of variations, rather than defaulting to the maximum out of curiosity.