Split Compound Text [LP]

Being straight with you up front: this is a node where the name promises more than the interface confirms. "Split Compound Text" strongly implies pulling a combined string apart into pieces, but what the node actually exposes is one text input and one text output - no separate output sockets for the parts, no delimiter field to tell it how to split. Neither the pack's README nor the node's own schema documents the exact rule it applies, so rather than guess at internal logic and risk telling you something wrong, here's what's actually verifiable and how to find out the rest yourself.

What's confirmed

It sits in the pack's Text category, takes a multiline string in, and returns a single string out under the output name text TEXT. Whatever transformation happens between those two - separating a "compound" piece of text into components and likely rejoining them in some cleaned-up form, given there's only one output - happens inside the node without any exposed parameter to control it.

Inputs and outputs that matter

• text (multiline, defaults to empty) - the string going in. That's the entire input surface.

Output: text TEXT - a single string. Not a list, not multiple sockets, despite the "split" in the name.

How to actually figure out what it does

Given the schema doesn't specify the splitting rule, the fastest honest path is to test it directly: drop the node, wire in a Show Text node on its output, feed it a few deliberately "compound" strings - a sentence with multiple clauses, a comma-separated list, text with obvious repeated structure - and watch what comes back. That'll tell you more in two minutes than any amount of guessing from the node's name.

Installing it

Through ComfyUI Manager, search "ComfyUI Level Pixel" or "ComfyUI-LevelPixel." Or:

cd ComfyUI/custom_nodes
git clone https://github.com/LevelPixel/ComfyUI-LevelPixel.git

Restart ComfyUI. No models or extra dependencies - it's a text-processing node like the rest of the pack's Text and Conversion categories.

Where people get burned

The obvious one: expecting multiple outputs from a "split" node and finding only one. If your workflow needs the pieces of a compound string as separate wires you can route independently, this node's current interface won't give you that - you're getting back a single processed string, not a list you can index into. If that's what you actually need, it's worth checking whether this node's output is meant to be fed into something else in the pack (a String Cycler, a further text node) rather than consumed directly as separated parts, and confirming with a real test string before building anything around an assumption about what "split" means here.