Nodes/ComfyUI_yanc/๐Ÿ˜ผ> Combine Channels
ComfyUI Node

๐Ÿ˜ผ> Combine Channels

Rebuild an image from three separate R/G/B channel images

By ALatentPlaceยทCreated 2 years agoยทUpdated 2 years agoยท 80
๐Ÿ˜ผ> Combine Channels
  • red
  • green
  • blue
  • image

This node only makes sense next to its missing-from-this-batch sibling: Divide Channels, also in ComfyUI_yanc, which splits an image into its red, green and blue components as three separate grayscale images. Combine Channels is the reverse - feed it three channel images, get one color image back.

How it works

Why would you split and recombine at all? Because sometimes the thing you want to edit is one channel, not the whole image. Level-adjust just the red channel to fix a color cast, sharpen only the green channel where most of the luminance detail lives, or use the split/edit/recombine loop to hand-build a texture that's really three unrelated grayscale renders packed into one RGB image (a cheap way to fake something like a packed ORM map). None of that is possible if you're always operating on full-color images - you need the channels isolated first, then a way to put them back together, which is exactly this pair of nodes.

The inputs and outputs that matter

  • red, green, blue (IMAGE, all required) - each slot expects the single-channel data for that channel, typically straight from Divide Channels' matching output.
  • image out (IMAGE) - the recombined result.

There's no intensity or blend control here - it's a direct channel pack, not a mixer. Whatever you feed each slot is exactly what ends up in that channel.

How to install it

Via ComfyUI Manager: search "yanc" or "ComfyUI_yanc" and install, then restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/ALatentPlace/ComfyUI_yanc

then restart ComfyUI. No extra dependencies or model downloads needed.

Common issues & troubleshooting

The recombined image looks tinted or wrong, not like your original. This almost always means you fed full-color images into one or more slots instead of true single-channel data from Divide Channels. If you skip the split step and just wire the same color image into all three inputs, you won't get your original back - you'll get whatever this node's channel-packing logic does with full RGB data in each slot, which isn't the intended use.

Shape mismatch or a broken-looking image. Make sure all three channel inputs are the same resolution. If one came from a resize step and the others didn't, you'll get a shape error or, worse, a silently misaligned result.

You don't actually need this node. If you just want to edit one channel and leave the others alone, and you're not doing anything exotic with the split channels individually, a simpler route is often a dedicated color-grading or curves node applied to the whole image. Combine Channels earns its keep specifically when you need per-channel independence that a normal filter node can't give you.

CategoryYANC/๐Ÿ˜ผ Image/๐Ÿ˜ผ Post Processing

Inputs (3)

NameTypeDefaultDescription
redIMAGEโ€”
greenIMAGEโ€”
blueIMAGEโ€”

Outputs (1)

NameTypeDescription
imageIMAGEโ€”