String Replacer

Cleaning up text is half the work of any automation: strip a model name out of a checkpoint path, normalize spaces, swap a tag separator, rewrite a prompt phrase before it hits the CLIP encoder. String Replacer does all of that in one node - with plain-string or regex matching, case-sensitivity control, and the bonus that it tells you how many replacements actually happened.

That count output is more useful than it sounds. It turns "did the replacement work?" from a visual check into a condition you can branch on.

How it works

You give it text, a search string, and a replace string. Plain mode does a direct substring replacement. Turn on use_regex and search is compiled as a regular expression with re.sub semantics, so groups and alternation work. Case sensitivity is a separate toggle - with it off, even plain-mode matching routes through a regex with re.IGNORECASE so the count stays accurate.

The count output is the number of replacements performed (re.subn's count), not a boolean. Zero means "nothing matched," which is your signal that something upstream changed.

Inputs

• text - the string to process (multiline allowed).
• search - what to find.
• replace - what to replace it with.
• use_regex - boolean, default off.
• case_sensitive - boolean, default on.

Outputs

• result - the modified string.
• count - how many substitutions happened.

Install

Standard pack install - ComfyUI Manager (search "ComfyUI-AutoFlow") or:

cd ComfyUI/custom_nodes
git clone https://github.com/ZXL-Xinram/ComfyUI-AutoFlow

Restart after. Needs only the stdlib re.

Where people get burned

• A bad regex returns the text unchanged. If the pattern fails to compile, the node logs the error and hands back the original text with count = 0. No crash - which is friendly but means a typo'd regex silently does nothing. Watch the console.
• count vs. match. A count of 0 is indistinguishable from "search found nothing" and "search didn't run." If you're branching on it, also branch on the text being non-empty.
• Regex metacharacters in plain mode are literal. That's the point of the toggle - but it means searching for a literal . requires use_regex off, or \. on. Know which mode you're in.
• Case-insensitive + regex is regex, so your search pattern still needs \b, \d etc. escaped correctly; the flag only changes case handling.

For prompt cleanup and filename normalization this is the most-workhorse node in the pack's string section. And the count output quietly gives your graph a way to know whether a transformation actually landed.