Loom Split

Here's the thing nobody tells you about c4f-wire-loom: the node you'll place first isn't Loom In or Loom Out, it's Loom Split - and it doesn't store or retrieve a single wire. Loom Split is a layout node. It takes one loom in and hands you three identical copies out, and its real job is giving you clean geometry to build your loom "bus" on. The pack's own tutorial starts with it: drop a few splits, link them into a chain, collapse them with their little arrows, and snap them into straight lines. Then you drag node slots onto the bus and ComfyUI's frontend auto-creates the matching collapsed In/Out nodes for you.

That workflow is the whole point of the pack. You end up with a tidy horizontal spine running through your workflow where all the "wires" are implicit, named labels in a bundle, instead of a hairball of spaghetti across the canvas. Think of it as the ComfyUI answer to bundling a dozen connections down one clean path - a loom-aware sibling of the rgthree Context wires, but one where you can also reach in, change a value, and put it back.

How it works

Mechanically it's almost insultingly simple: the loom is a Python dict of label→value, and Loom Split just copies it three times.

• loom (optional) - if nothing's connected, you get three empty looms, which is fine; it's inert.
• Outputs: three LOOM outputs, each an independent copy of the input.

Because every output is a copy, you can branch in three directions and each branch reads the full loom. It doesn't "split" wires in half - nothing gets cut. It's a fan-out tap. If that sounds redundant next to just connecting the same wire to three places, you're right, and that's the point: the split exists so you can collapse it into a straight line and make the graph readable. The copies matter for one practical reason though: each branch can feed a different Loom Join or Loom In and modify its own copy without scribbling on the other branches.

Where it fits

Use it the way the tutorial does - as the backbone. Place it first, chain several together, collapse them, and drag your node inputs and outputs onto the spine. The frontend recognizes the loom slots and auto-creates the right typed nodes (Loom Image In/Out, Loom VAE In/Out, and so on), which is where the real time saving lives. Alt-click on a split to clone it as you build the line out.

Install

Same as the rest of the pack - no requirements.txt, no models, no API keys:

cd ComfyUI/custom_nodes
git clone https://github.com/Codes4Fun/c4f-wire-loom

then restart ComfyUI, or install c4f-wire-loom from ComfyUI Manager.

Troubleshooting

• Splits that don't look collapsed - the collapse toggle is the arrow in the node's top-left; in ComfyUI's Nodes 2.0 frontend the README warns that dragging a collapsed node by that same arrow can also toggle it back open. The author shipped a workaround in the frontend code, so make sure the pack is up to date.
• Branches not getting a wire - check that the Loom In feeding the split actually stored the wire before the split in graph order. If the In is downstream of the split, the copies never contained it.
• Forgetting to chain splits - each split is its own tap. If your third branch is wired to a split that isn't connected to the first, it's reading an empty loom. Follow the tutorial's chain and you're fine.