Greyscale Noise

Here's the difference between JDC_RandNoise and JDC_GreyNoise: random noise rolls three independent values per pixel, one per channel. Greyscale noise rolls one value per pixel and then maps it through each channel's range. So where the random node produces colorful static, this one produces neutral, monochrome texture - which you can then tint however you like with the per-channel controls.

It's part of the noise family in comfy-plasma by Jordach, and it's the one you want when you care about the shape of the noise more than its color.

How it works

The README's example explains the mechanism better than any abstract description: a value of 128 at a pixel, mapped with red set to work between 0 and 128, produces 63 in the red channel. One base random value, then a linear remap into each channel's range. Leave all the per-channel ranges at their -1 default (meaning "use the global value range") and all three channels get the same value - a true greyscale noise image.

The moment you override one channel's range, you get a tinted wash. Want warm noise? Nudge red up and blue down. Want cold static? Flip it. The base texture stays monochrome-coherent, which is the whole point - the channels are correlated, not independent, so the result reads as "texture with a color cast" rather than "rainbow static."

The standard family controls apply: width / height (128–8192, steps of 8), value_min / value_max (default -1 = full 0–255 range), per-channel red/green/blue_min/max overrides, and a seed.

When you'd reach for it

• Neutral noise for masks and overlays where color would get in the way.
• A texture base you want to tint subtly without doing channel math yourself - the per-channel ranges are the channel math.
• Film-grain-style overlays with the color left to a later grading step.

It's the polite middle child of the noise family: calmer than random, more controllable than pink or brown.

The seed caveat

Same story as every node here: the seed makes output reproducible, but the author's source comment warns it "does nothing" in terms of ComfyUI's caching - you have to change the seed widget (or another input) to force a fresh roll. There's no auto-randomize.

Install

No requirements.txt, no models - the pack only needs Pillow, numpy, torch, and requests, all already in ComfyUI.

ComfyUI Manager → Install Custom Nodes → search comfy-plasma → install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/Jordach/comfy-plasma

Restart ComfyUI. It's under image/noise.

Common issues

The usual noise-node cautions apply: it's a Python per-pixel loop, so it's snappy at 512 but a real wait toward 8192 - generate small and upscale. And one easy confusion: because all channels inherit the same base value by default, the output can look "too flat" compared to random noise. That's not a bug - it's greyscale. If you expected colorful static, you've grabbed the wrong node; JDC_RandNoise is next door.