List from Loop (PixNodes)

A loop that produces something every iteration is only useful if you can collect what it made. List from Loop is the collector: it accumulates each iteration's output into a growing list (or a proper image/latent batch), so that when the loop finishes you have one tidy result - every generated image, every prompt, every JSON object - ready for downstream use.

The name says where it lives: this node sits inside a loop, built for the pack's For/While loop family (the PixNodes/loop category, with flow control ports). The loop-start node hands you an index and values; you run your per-iteration work; this node appends that work to the collection; you feed the collection back around the loop; and the accumulated result comes out at the end.

How the accumulation works

Three inputs: list_in (the collection from the previous iteration, or the empty start), item (the new data this iteration produced), and split_items (boolean, default on). split_items controls whether a list item gets spread into the collection element-by-element (extend) or appended as a single nested entry. If a downstream step hands you a whole list in one iteration and you want each element kept separate, leave it on.

The smart part is what happens at the end. If everything collected is a tensor with matching shapes, it's torch.cat'd into a single proper image batch - so collecting images across loop iterations gives you a real IMAGE batch you can feed a sampler or save, not a Python list of tensors. The same trick applies to latents: collected {"samples": ...} dicts get their tensors concatenated, noise masks included when they align. Strings and numbers just accumulate as a plain Python list.

Outputs are collection (the accumulated result) and length (INT), the running count - handy as a loop bound.

Install

Part of Comfyui-PixNodes. ComfyUI Manager → search "PixNodes" → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/pixixai/Comfyui-PixNodes

then restart. (The README's manual-install line pastes the wrong repo URL, ComfyUI-AlignLayout - use the one above.)

The loop gotchas

Loops are the fiddliest thing in this pack, and the changelog is the proof: v1.1.0's entire entry is "fixed loop node data not transferring." If a loop looks like data isn't arriving at list_in, update the pack first. Two habits keep you out of trouble: always feed list_in from the loop's previous round (not a fresh node), and remember the tensor-cat only fires when shapes match - collect images of different sizes and you'll silently get a Python list back instead of a batch, which downstream IMAGE ports will reject. Keep the per-iteration output uniform and this node stays the boring, reliable collector it's meant to be.