Image Batch to List 🐼

What it is

A small plumbing node that solves a specific ComfyUI annoyance: batches and lists of images aren't the same thing, and a lot of nodes only accept one or the other. A "batch" is a single IMAGE tensor with multiple images stacked along a batch dimension - what you get out of most samplers. A "list" is several separate IMAGE outputs treated as individual items, which is what nodes built around per-item looping (and this pack's own ImageRewardScore-fofo, for instance) expect. Image Batch to List's whole job is taking up to four separate batches and merging them into one list output.

If you've ever tried to wire a batch straight into a node that wants a list and gotten a shape mismatch or only the first image processed, this is the node that fixes it.

How it works

There's no image processing happening - no resizing, no compositing. It just takes each batch input, splits it into its individual images if it isn't already, and concatenates everything into a single list. Four batches in, one combined list out.

Inputs and outputs

• image_batch1 - required, the first (and possibly only) batch.
• image_batch2, image_batch3, image_batch4 - optional. Leave them unconnected if you've only got one source; wire them up if you're merging images from multiple generation branches (say, outputs from two different samplers you want scored or saved together) into one combined set.

Output is IMAGES, list-typed - every image from every connected batch, flattened into one list.

Installing it

It ships as part of the Comfyui_image2prompt pack, alongside the captioning and prompt-generation nodes. Via ComfyUI Manager, search "Comfyui_image2prompt" and install. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/zhongpei/Comfyui_image2prompt

Restart ComfyUI afterward. (The README's own example command has a typo - it clones a hyphenated URL that doesn't actually resolve; use the underscored one above.) No model downloads are involved for this particular node - it's pure Python, no weights to fetch.

Common issues

There isn't much surface area for this one to break on its own - it's a simple merge utility, not a model call. The failure modes that show up in practice are more about why you reached for it in the first place:

Wiring a batch into a node that wants a list, or vice versa, without this conversion. If a downstream node silently only processes the first image, or throws a dimension error, check whether it actually wants a list rather than a batch - this is the node that bridges the two, and it's easy to not realize the mismatch exists until you hit it.

Leaving inputs unconnected isn't an error, but skipping this node entirely when you don't need multiple sources is fine too - if you only have one batch to convert, you can wire just image_batch1 and ignore the rest; the optional inputs really are optional.

General pack-install friction. As with every node in this pack, if it doesn't appear in your node list at all, the first thing to check is the ComfyUI console log at startup - a dependency conflict elsewhere in the pack (it shares an install with several heavier VLM and LLM nodes) can prevent the whole pack from loading, this simple node included.