Image Batch to List

Every ComfyUI user eventually hits the wall where one node hands you an IMAGE batch and the next node wants an image list. Those are two different things in this ecosystem - a batch is one tensor with a B dimension, a list is a Python list of separate tensors - and not every pack supports both. Image Batch to List is the three-line bridge that converts one to the other.

It's part of ComfyUI 1hewNodes (author: 1hew, a solo dev whose pack reads like a personal workflow toolbox - bilingual README, active v3.x changelog, minimal community footprint). The node itself is about as simple as a node gets, which is exactly its job.

How it works

Under the hood it takes your [B,H,W,C] tensor and slices it into B separate [1,H,W,C] tensors, returned as a list. That's it. No resizing, no normalization, no cleverness - each item keeps its own size, so you can even pipe in a batch where frames differ in resolution and the list items stay independent.

The reverse direction is covered by the pack's Image List to Batch node, so the two form a round trip: batch → list → do per-image work → back to batch.

The inputs and output

• image_batch - the IMAGE batch you want to split apart.

Output is image_list (an IMAGE-typed list, one entry per frame). Wire it into any node that declares an image-list input - per-frame processors, list-based iterators, samplers that want individual items. If you're looping over frames "one at a time," this is the standard way to get there.

Installing it

ComfyUI Manager → search "1hew" or "ComfyUI 1hewNodes" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/1hew/ComfyUI-1hewNodes

The pack's requirements install opencv, scikit-image, scikit-learn and heavier detection deps (ultralytics, rembg) even though this conversion node only needs numpy. First install is slow; that's the price of the grab-bag. No models downloaded for this node.

Gotchas

• Empty batch - the node returns an empty list instead of crashing, which is polite but means anything downstream that assumes a minimum length will throw. Guard with a check if you feed it variable-length batches.
• Update ComfyUI first - the whole pack is built on the current node API (comfy_api.latest), so on a stale build the node won't register. A "node class not found" error here usually means update ComfyUI, not debug the pack.
• It's a dumb converter on purpose - if you need size-normalized frames while converting, that's what Multi Image Batch or the resize nodes are for. Keep this one dumb.