String Splitter (CRT)

Storyboard-style workflows are built on a simple idea: write all your shot descriptions in one text box, one paragraph each, and feed each paragraph to its own generation. The boring-but-essential step in the middle is splitting that blob of text into discrete strings you can route to different samplers or conditioning. That's this node. It splits a single multiline string into up to 64 separate STRING outputs, either on blank lines or on a custom separator you provide.

It looks intimidating because the node always shows a wall of outputs - string_1 through string_64. Don't panic. You only use the ones you set, and a split_count widget controls how many are active.

How it works

Default behavior: it splits your text on empty lines (one or more blank lines act as the boundary), so each paragraph becomes its own output. If you set split_symbol to something like --- or |, it splits on that instead - handy when your paragraphs themselves contain blank lines for formatting.

Two behavioral details are worth internalizing:

• strip_whitespace (default on) trims each part so you don't get stray newlines gumming up your prompts.
• Empty parts are dropped entirely. Split by blank lines, and two blank lines in a row won't hand you an empty string - the parts are compacted.

The split_count input (1–64) says how many of those outputs you're actually using, which also keeps unused outputs from carrying stale values from a previous run.

Inputs and outputs

• text - the blob to split (multiline).
• split_count - how many outputs are active.
• split_symbol - custom separator; empty = split on blank lines.
• strip_whitespace - clean up each part.

Outputs are string_1 through string_64, each a STRING. Feed the first few into your prompt encoders and leave the rest unconnected.

Installing it

CRT-Nodes install as usual: ComfyUI Manager → CRT-Nodes, or:

cd ComfyUI/custom_nodes
git clone https://github.com/PGCRT/CRT-Nodes.git
pip install -r requirements.txt

Restart; it's under CRT/Text.

Where people get burned

The numbering is the trap. Because empty parts are dropped, part N of your input is not guaranteed to be output string_N - a leading blank line or a double blank line shifts everything by one, and if you've hardwired paragraph three to string_3 you'll silently get paragraph two. If your outputs are misaligned, count your blank lines. Also, keep split_count honest: it's easy to leave it at 2 while you've written five paragraphs and wonder why three are missing - bump it and reconnect.