What It Does

Word processors and chat apps quietly rewrite your punctuation. Straight quotes become curly ones, double hyphens become em dashes, three dots become a single ellipsis glyph, and ordinary spaces become non-breaking spaces. It looks fine in the original app and causes problems everywhere else: broken code samples, noisy diffs, mismatched search results, and characters that render as boxes in monospace.

This tool normalizes all of it in one pass:

Curly quotes → straight quotes, including the single and double variants
Em and en dashes → hyphens
Ellipsis character → three dots
Non-breaking, thin, and ideographic spaces → a regular space
Zero-width and bidi control characters → removed
Spacing cleanup - space before punctuation, repeated spaces, trailing whitespace, and line-ending normalization

Every Rule Is A Toggle, With A Count

Fourteen rules, each independently switchable, each reporting how many replacements it made. That matters: a silent bulk-replace tool is one you cannot trust with a long document. If you want curly quotes preserved but em dashes gone, switch off one rule. The counts tell you at a glance whether the text was as messy as you thought.

Sensible defaults are on; the more opinionated rules (primes, bullet characters, collapsing blank lines) are off until you ask for them.

What It Is Not

This is a typography tool - it normalizes characters. It is worth being precise about one thing, because the two get conflated: it does not remove statistical AI watermarks like Anthropic’s Claude mark or Google’s SynthID-Text. Those signals live in which words a model chose, not in punctuation, so no character-level cleanup touches them. If you want to know what is genuinely detectable in a piece of text, the AI Provenance Inspector covers it honestly.

Pairs with the Text Formatter for case and line operations, and the Hidden Text Inspector when you want to see hidden characters before removing them.

Limitations

It does not know your intent. Converting curly quotes to straight ones is correct for code and plain text, and wrong for typeset prose. Switch off what you do not want.
Apostrophe direction is not inferred - every curly apostrophe becomes a straight one regardless of position.
Language-specific punctuation (French guillemets, German low quotes) is normalized to ASCII quotes, which may not be what you want for published prose in those languages.