Images 2 Image

A lot of ComfyUI nodes give you a batch of images and expect you to keep treating them as a batch. Sometimes you don't want that - you want image #1 going to one place and image #3 going somewhere completely different. Images2Image is the un-batching node: hand it a batch, get back up to five separate single-image outputs you can wire independently.

How it works

It takes one IMAGE input - which, in ComfyUI, is really a batch (a stack of one or more images) - and splits it out positionally into five distinct outputs. Nothing about the pixels changes; this is purely about routing.

The inputs and outputs that matter

• images - the batch coming in.

Outputs: image1 through image5 - each one a single IMAGE, pulled out by position from the incoming batch. Wire each to wherever that specific image needs to go - separate save nodes, separate captioning calls, separate downstream edits, whatever your workflow calls for.

If your incoming batch has fewer than five images, the schema doesn't specify what the unused outputs carry - treat outputs beyond your actual batch size as undefined rather than assuming they'll be blank or duplicated, and check what you actually get before building critical logic on top of them.

How to install it

Via ComfyUI Manager: search comfyui_LLM_party, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/heshengtao/comfyui_LLM_party.git

Then pip install -r requirements.txt from inside the pack folder using ComfyUI's own Python, and restart. Nothing extra needed for this specific node - no models, no API keys, it's pure image plumbing.

Common issues & troubleshooting

One or more outputs come back empty or wrong. Almost always a batch-size mismatch - if you sent in 3 images but are relying on image4/image5, those weren't in your original batch to begin with. Check the actual size of what's feeding into images first.

You only needed to split off one image, not five. This node always exposes all five outputs regardless of how many you actually use - that's fine, unused outputs just don't get wired to anything downstream. There's no smaller variant; if five is overkill for your case, you can simply ignore the rest.

Order doesn't match what you expected. The split is positional - whatever order the images sit in inside the incoming batch is the order they come out as image1...image5. If ordering matters and it's coming out wrong, the fix is upstream, in whatever produced the batch, not in this node.