← All tools / Style string cleaner
Style string cleaner
Paste a Style line that grew out of hand. Get back one clean, comma-separated list: duplicates gone, hype words stripped, spacing fixed, and trimmed to a character budget if you set one.
Starting from scratch instead? Use the Prompt Builder. Want the opposite list? Exclude styles generator.
Style string
Waiting for a Style string
A budget of 0 means no limit. When a budget is set, whole tags are dropped from the end until the line fits — partial tags are never left behind.
What the cleaner actually changes
- Splits on real separators. Commas, semicolons, pipes and line breaks all count as tag boundaries, so a string pasted out of a spreadsheet or a notes app still parses.
- Removes duplicates case-insensitively. "Dream Pop" and "dream pop" are the same instruction; the first one wins.
- Strips hype vocabulary. Words like amazing, viral, best and masterpiece describe how you feel about the result, not what the music should do. They're removed from inside tags; what's left of the tag stays.
- Normalises whitespace and empty slots. Double commas, stray trailing separators and runs of spaces disappear.
- Applies a budget by whole tags. Truncating mid-phrase creates nonsense instructions, so tags are dropped from the end instead.
Order matters, so keep it by default
Leave "Keep original order" on unless you have a reason not to. Earlier tags carry more weight, and re-sorting alphabetically will quietly promote whatever happens to start with an A. The alphabetical mode is there for auditing a long list, not for generating.
Reading the report
The panel under the options tells you how much actually changed: characters before and after, how many duplicate tags were dropped, how many tags had hype words removed, how many collapsed to nothing, and how many were cut to meet your budget. If the "cut for budget" number is large, your list is longer than the budget can carry — go back to the builder and decide what to lose deliberately.
A sensible workflow
Draft in the builder, clean here, then keep the cleaned line as your baseline for testing. When a generation goes wrong, change one tag in the baseline rather than rewriting the whole string — otherwise you can't attribute the difference to anything.