调整饱和度 / Saturation Adjustment

Saturation Adjustment is the simplest useful node in the RUI-Nodes pack, and honestly one of the easiest wins you'll add to a graph. It takes an image and a single number and gives you back the image with more or less color. That's the whole job. No model downloads, no API key, no hidden behavior - you could write the thing in twenty lines, and RUI-Nodes did, which is exactly why it's nice to have around.

You'll reach for it in three situations. Converting a color render to black-and-white without fighting with color-math nodes. Dragging a too-vivid generation back down to something that doesn't hurt to look at. And adding a touch of punch to a flat image before it goes through an upscaler. It's the kind of node you drop into a workflow once and forget it's there.

How it works

Under the hood it's PIL's ImageEnhance.Color, wrapped to accept ComfyUI's standard IMAGE tensor (batch, height, width, channels, float 0–1). Each frame of the batch gets converted to a PIL image, enhanced, and converted back. That means it's genuinely batch-friendly - feed it a 16-frame batch and it processes every frame, not just the first one.

The one knob that matters:

• saturation - 0.0 is fully desaturated (true black-and-white), 1.0 leaves the image untouched, and anything above 1.0 cranks the color up. The range goes to 5.0, but the author's own tooltip warns that past roughly 2.0 colors start blowing out and looking fake. Treat 1.0–2.0 as the usable sweet spot for "more vivid," and 0.0 for "monochrome."

Output is a single IMAGE tensor, same shape as the input, ready to wire into whatever comes next.

Installing it

This ships as part of RUI-Nodes (display name "Rui-Node🐶"), a bilingual Chinese/English pack that bundles a couple dozen image, text, mask, and AI-API nodes. Install once and every node in the pack - this one included - appears. Easiest route is ComfyUI Manager: search "RUI-Nodes" and hit install, then restart. Or do it by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/rui40000/RUI-Nodes
cd RUI-Nodes
pip install -r requirements.txt

Then restart ComfyUI. The base requirements are just torch, numpy, Pillow, and requests - all things your ComfyUI install almost certainly already has, so this node needs nothing extra.

Where people get burned

Two things, both minor. First, the node routes everything through PIL's 8-bit pipeline, so you lose a little precision on the round-trip - invisible for practical purposes, but if you're chaining a dozen of these it's worth knowing. Second, because it silently processes the whole batch, it's easy to forget it's in the graph and wonder why your output looks slightly different from the preview. It isn't a trap; it's just quiet.

If you only ever want grayscale, saturation = 0 is the whole trick. Everything else is extra credit.