Images Convert RGB

This is the pack's boring node, and boring is the compliment. It takes an image (or a batch of images) and converts it to RGB - meaning it drops the alpha channel. That's it. No parameters, no cleverness. You'll want it the first time a node hands you an RGBA image - a VAE decode of a latent with transparency, the output of a background-removal node, a PNG loaded with an alpha layer - and the next node in your graph chokes with a channel-count error or silently treats the transparency as black. ComfyUI's IMAGE type doesn't care whether you feed it 3 or 4 channels, but a lot of nodes very much do.

How it works

Each frame in the batch is converted with Pillow's .convert('RGB'), which just drops the alpha component and composites nothing - it doesn't flatten against a background, it discards the channel. There's no blending or color math involved. The one thing worth knowing: it's batch-aware. The code splits the input into individual frames, converts each, and concatenates them back into a single tensor - so a batch of 16 RGBA frames comes out as a batch of 16 RGB frames, no indexing, no ImageBatch juggling.

Inputs and outputs

• images - the only input, an IMAGE tensor, single frame or batch.

Output: one IMAGE, same batch count, 3 channels.

Installing it

Same pack as everything else here:

cd ComfyUI/custom_nodes
git clone https://github.com/AhBumm/ComfyUI_UnfakePy_Warpper

or ComfyUI Manager → Install Custom Nodes → search ComfyUI_UnfakePy_Warpper, then restart. It shares the pack's requirements - unfake==1.0.7 and nest_asyncio - and Python 3.10+; no models, no GPU.

When to reach for it

Drop it in line any time a downstream node is doing something weird with an image that might have an alpha channel: color conversions, image-to-latent encodes, anything that assumes three channels. It's also the natural companion to the pack's own Remove Background (FloodFill) node when you want the cutout for inspection but the actual RGB pixels for the next step. A one-purpose utility node isn't glamorous, but it's the kind of thing that saves you ten minutes of "why is everything a rectangle" when you're mid-workflow. If you never need it, you never missed anything - but you'll be glad it's there the first time you do.