RGB to Greyscale

JDC_Greyscale ("RGB to Greyscale") does the obvious thing - it converts a color image to black and white - but it does it in a way that's easy to miss and genuinely useful: the output is still an RGB image, just with all three channels equal. That means you get the monochrome look without breaking every downstream node that expects a 3-channel tensor.

It's the color-stripping utility in comfy-plasma by Jordach, and it's the kind of node you don't think you need until a workflow keeps failing on channel counts.

How it works

Mechanically it's a quick trip through Pillow: the image is converted with ImageOps.grayscale (standard luminance weighting - the human-eye-correct "0.299R + 0.587G + 0.114B" style mapping, not a naive average), and then pasted into a fresh RGB image so it comes back out with three identical channels.

Why does that matter? A lot of ComfyUI nodes are written assuming 3-channel RGB input. If you naively convert to a 1-channel greyscale image, nodes that expect [batch, height, width, 3] will throw shape errors, and you'll be googling tensor-size stack traces at 11pm. By keeping the output RGB with equal channels, this node lets you strip color without triggering any of that. It's a compatibility-conscious choice that saves real pain.

Input: IMAGE. Output: IMAGE, still 3-channel.

When you'd reach for it

• Making a monochrome version of a render to compare composition without color distracting you.
• Feeding a greyscale base into an img2img pass where you want color to come from the prompt, not the source.
• Prepping line-work or texture work where hue would be noise.
• Before JDC_EqualizeGrey if you want luminance-only equalization without the per-channel color shift - greyscale first, equalize, then you've got a clean monochrome result.

Install

No requirements.txt, no models. 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/postprocessing.

Common issues

The one thing to be aware of: this is a destructive conversion in the sense that once color is gone, it's gone - there's no "desaturate by X%" here, just full monochrome. If you wanted a partial desaturation, this isn't the node. And because it keeps RGB channels equal, the file on disk will be bigger than a true single-channel greyscale save would be; if you're optimizing storage, do the greyscale conversion at save time instead. Otherwise it's a boring, reliable node that does exactly one thing and gets out of the way.