Text Replacer

This is the node in the String Utils pack with actual teeth. The formatters and converters are plumbing; Text Replacer is where you clean up, transform, and restructure text - and the mode switch to regex is what makes it more than a find-and-replace. If you've ever wanted to strip a tag out of every prompt, swap a character name across a workflow, or do a pattern-based rewrite, this is the node.

What it does

Three text inputs and a mode selector. Give it the source text, a replace pattern, and a replacement, and it returns the transformed string.

• normal mode - literal string replacement. text.replace(replace, replacement). "John" → "Mr. Smith", everywhere it appears.
• regex mode - re.sub(replace, replacement, text). The replace field is a regular expression, and replacement can use backreferences to captured groups.

The README's regex example is the one to internalize: with replace="([0-9]+) is" and replacement="\1 is in fact", the text "42 is the answer" becomes "42 is in fact the answer" - the \1 pulls the captured 42 into the replacement. That pattern - match a chunk, keep part of it, swap the rest - is genuinely powerful for prompt cleanup.

The inputs that matter

• text - what you're transforming.
• replace - what to find: a literal string in normal mode, a regex pattern in regex mode.
• replacement - what to put in its place.
• mode - dropdown, normal (default) or regex.

Output: text, the transformed STRING.

How to install it

Part of exectails/comfyui-et_stringutils ("String Utils"). ComfyUI Manager → search "String Utils", or:

cd ComfyUI/custom_nodes
git clone https://github.com/exectails/comfyui-et_stringutils

Restart, and it's under exectails/Strings. No dependencies, no model downloads - the only library it touches is Python's built-in re module.

Common issues

• Normal mode is literal. Put "(" in the replace field in normal mode and it looks for a literal parenthesis. Regex metacharacters only act like patterns in regex mode. If your text contains regex characters and you're not intending a pattern, stay in normal mode.
• Invalid regex = crash. A malformed pattern (say, an unclosed () raises an error and kills the run. Test your pattern elsewhere first if it's gnarly.
• No matches is not an error. re.sub and str.replace both return the text unchanged when nothing matches - the node just passes it through. That's usually what you want, and it means a missing tag in your data won't take the whole workflow down.
• Backreference syntax. \1 works in the replacement for regex mode (that's what the README shows), and \g<1> works too. Note those backslashes are meaningful only in regex mode - in normal mode \1 is a literal backslash-one.

Use it to strip stray tags, normalize whitespace, swap names, or build a regex-powered cleaning stage between your prompt source and the encoder. It's the utility node that quietly saves you from editing prompts by hand.