About plainkit.tools
plainkit.tools is a free text cleaner for pasted text. It removes invisible Unicode characters and normalizes LLM-flavoured typography to plain, portable ASCII, entirely in your browser. There is no account, no upload, and nothing to install.
What it does
Text copied out of chat tools, word processors, web pages, and PDF exports carries characters you cannot see: zero-width spaces, joiner residue, byte-order marks, non-breaking spaces where normal spaces belong. It also carries lookalike punctuation: curly quotes, long dashes, ellipses, bullets, arrows, fake-bold math letters, and fullwidth forms. These break search, diffs, JSON parsers, and code in ways that look like mystery failures, because the text appears correct on screen. The cleaner scans pasted text, highlights every issue in place with its code point, and produces a clean version with visible words unchanged. Code wrapped in backticks passes through byte-identical.
How it works
The cleaning is done by detypo, an open-source engine with zero runtime dependencies. One pipeline inspects the text and reports issues with exact offsets; the same report drives both the highlighted view and the cleaned output, so what you see marked is exactly what gets changed. Context-sensitive rules keep characters that carry meaning: joiners inside emoji sequences and Indic or Arabic script shaping are preserved, while the same code points floating between Latin letters are stripped. The web app runs that engine locally on the page. Pasted text never leaves the device, which is also why the tool works offline once loaded and why there is nothing on a server that could leak.
What it is not
This is a copyediting tool, not a watermark remover. It normalizes visible typography and strips invisible Unicode carriers. It does not defeat statistical or token-sampling watermarks, which live in a model's word choices rather than in characters that can be located and removed, and it does not touch file provenance markings such as C2PA, XMP, or EXIF metadata. The boundary is deliberate: normalizing punctuation is copyediting, while circumventing transparency markings is a different category, and the tool stays on the copyediting side. Detectors that judge how writing reads are unaffected by cleaning, and no claim to the contrary is made anywhere on this site.