The text is lowercased, symbols with a spelling are written out (& becomes “and”, ß becomes “ss”), accented letters are Unicode-normalised so the accent can be dropped, and every run of spaces or punctuation collapses into a single separator. Apostrophes inside words are removed rather than replaced, so “don’t” becomes “dont”, not “don-t”. Stopword removal and the length limit are applied last, and the “what changed” list records each step that actually fired so the output is never a surprise.
Everything runs in your browser — nothing you type is sent to a server.