Encoding Fixer

How to fix broken accents in text

Paste it on the left. The repaired text appears on the right immediately, and there is nothing to configure.

If nothing changes, the text was not mis-encoded and the panel says so. That is worth knowing on its own, because it means the problem is somewhere else.

Why é turns into é

Because two programs disagreed about how to read the same bytes.

In UTF-8 the letter é is stored as two bytes, C3 and A9. A program that reads one byte at a time using Windows-1252 sees those as two separate characters, Ã and ©, and shows both. Nothing was corrupted. The bytes are exactly what they always were, and only the reading was wrong.

That is why the damage is so uniform. Every accented letter becomes two characters, and every curly quote, dash or ellipsis becomes three, because those are three bytes in UTF-8. Once you see ’ a few times you can read it as a right single quote without thinking.

Where it comes from

Exporting a CSV from Excel is the classic. Excel writes UTF-8 without a byte order mark, then reads it back as the system code page, so a file survives one round trip and breaks on the next.

The rest are the same story with different names: a database column declared latin1 holding UTF-8 data, an old CMS import, a shell pipeline missing a locale, an email header. In each case the bytes moved correctly and the label travelling with them did not.

Why it can always be undone

Because mojibake loses nothing. The repair is the original mistake run backwards: turn each character back into the byte it stood for, then decode those bytes as UTF-8.

The check is built into that. If the bytes are not valid UTF-8, the text was never mojibake and nothing is changed. There is no confidence score and no threshold, which is why this tool has no settings.

One case genuinely cannot be recovered, and it is worth knowing which. If the bad characters were already replaced by question marks or by the diamond shaped replacement character, the bytes were thrown away at that point and no tool can bring them back. Go and re-export the file.

Text mangled more than once

A file that already shows é and gets exported wrong a second time becomes é. Most fixers undo one layer and hand back something that still looks broken.

This one repeats until a pass stops changing anything, so double and triple encoding come out clean and the panel tells you how many layers there were.

What to check next

Repaired text often still carries the curly quotes, em dashes and non-breaking spaces that were in the original, since those survive the round trip perfectly well. The AI text cleaner strips those in one pass if the text is heading somewhere that cannot take them.

If a specific line still refuses to import after all that, the invisible character detector shows what is actually in it, character by character, with a line and column.

Everything runs in your browser. Nothing is uploaded, which matters for the exports people usually bring here.