Multi Image List

In ComfyUI, "I have several images" can mean two different data shapes, and confusing them is a classic beginner wall. A batch is one tensor with a frame dimension - [B, H, W, C]. A list is an actual Python list of separate tensors. Some nodes only accept the latter, and you don't get one from the former just by squinting at it. Multi Image List converts batch(es) into a proper list, one image per element, ready for list-expecting nodes.

What it's for

You'll want it the moment something hands you a batch and the next node only eats lists - per-image processing nodes, custom iteration, nodes that index into image_N style inputs. It also accepts multiple input wires: image_1, image_2, … each of which can be a batch or a list, and the node flattens everything into one ordered list. So it doubles as a lightweight "merge all my image sources into one stream" node.

How it works

Each incoming value is checked: a single image becomes a one-item list item, a batch gets split frame-by-frame into individual [1, H, W, C] tensors, and a list gets recursively flattened. Order is preserved - inputs are collected in wire-number order, and within a batch, frames stay in order. Values that are empty or not image tensors are skipped silently, so a disconnected input doesn't explode the graph.

The distinguishing feature: this is a list-output node, flagged in its schema so other list-aware nodes know they're talking to a list. That matters because some of the pack's companion nodes and various community nodes specifically test for list vs. batch inputs to decide how to iterate.

The inputs that matter

• image_1 (IMAGE, optional) - plus dynamic image_2 … image_N wires that appear as you drag more in.

One output:

• image_list (IMAGE, as a list) - every frame as its own item.

Installing it

ComfyUI Manager → search ComfyUI-1hewNodes → install → restart, or:

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

Restart ComfyUI. No model downloads. The pack's requirements pull heavyweight packages (rembg, ultralytics, transformers) even for nodes that don't need them, and it targets ComfyUI's newer comfy_api.latest API - update ComfyUI if nodes fail to register.

Common issues

The output being a list is a feature and a trap: a node that does accept batches won't necessarily accept this list, and vice versa. If a downstream node complains about a type mismatch after this node, that's the list-vs-batch split showing up - that node wants the batch directly, so skip this node or add the pack's batch/list converter on the other side. Also note items are output as singleton [1, H, W, C] tensors, so if a downstream node wants flat 3D frames, you may need to squeeze along the way.