Turn “café”, “don’t” and “привет” back into real text. Reverses UTF-8 misread as Windows-1252 or Latin-1, even when it happened twice or only in parts of the text. Runs in your browser.
Pasted text has already lost bytes when it shows ? or �. If you still have the .csv / .txt / .srt file, pick it here: each common encoding is tried on the raw bytes and ranked by how plausible the result looks.
Limits: characters that were already replaced by “?” or “�” cannot be recovered from text alone. Encoding guesses are heuristics — read the preview before relying on a result.