Text Cleaner

The problem this solves

Text copied out of a PDF or an OCR scan arrives with the shape of the page still baked into it. Every visual line ends in a line break, words are split with hyphens across those breaks, bullets have become stray characters, and justified text has left double spaces everywhere. Paste that into a document and it reflows into nonsense.

Each of the fixes below is mechanical, which is exactly why it should not be done by hand.

What each option does

• Join broken line wraps — removes a line break when the line did not end in punctuation and the next line starts in lower case. That is the signature of a wrapped line rather than a real paragraph break, so paragraphs survive.
• Rejoin hyphenated words — turns hyphen-\nated back into hyphenated. Applied before line joining, since it is the more specific case.
• Collapse double spaces — reduces runs of spaces and tabs to one. Common in text lifted from justified columns.
• Strip invisible characters — removes zero-width spaces, soft hyphens, word joiners and byte-order marks, and converts non-breaking spaces to ordinary ones. These are the characters that make two identical-looking strings refuse to match.
• Straighten curly quotes — converts typographic quotes and apostrophes to straight ones. Off by default, because in prose you usually want to keep the curly ones. Turn it on for code, CSVs and configuration.
• Normalise dashes — em dashes to -- and en dashes to -. Also off by default, for the same reason.
• Tidy bullet characters — turns •, ●, ▪, ·, ‣ and ◦ at the start of a line into a plain -, which most editors then recognise as a list.
• Trim trailing spaces — removes whitespace at the end of every line.
• Collapse blank lines — reduces three or more consecutive newlines to a single blank line.

A good order of work

1. Paste the raw text and leave the defaults on.
2. Read the first two paragraphs of the result. If real paragraph breaks have been swallowed, switch off Join broken line wraps and fix the few that need it by hand.
3. Only then turn on the quote and dash options, and only if the destination is code or data.

Common questions

Is my text uploaded?

No. Every transformation runs in your browser on this page.

Will it fix OCR mistakes?

No — it fixes layout artefacts, not recognition errors. A misread word is still a real word, so nothing can flag it automatically. There is more on this in OCR, explained without the jargon.

Can I get the original back?

Yes. The input box is untouched; the cleaned version appears separately, so you can always compare or start again.