StringMash.com

Mojibake fixer

Paste text full of ’ and é, and get back the apostrophes and accents it should have had.

71 characters
Updates as you type
Correct text is left as it is

Show the steps
  1. Each garbled run was turned back into the bytes Windows-1252 showed, then read as UTF-8, as it should have been.
  2. ’ → ’ (U+2019): bytes E2 80 99
  3. — → — (U+2014): bytes E2 80 94
  4. é → é (U+00E9): bytes C3 A9
  5. “ → “ (U+201C): bytes E2 80 9C
  6. â€[U+009D] → ” (U+201D): bytes E2 80 9D
Printable chart

Using the fixer

Paste the garbled text and copy the result. Every broken sequence is repaired at once, and anything that was already right stays exactly as it was, so you can paste a whole document that's only partly broken. The working lists each repair, with the bytes behind it and a link to the page about that sequence.

Text that was garbled twice, so ’ became ’, is undone layer by layer. Three switches tidy up afterwards if you want them: remove a byte order mark, turn no-break spaces into ordinary ones, and straighten curly quotes. All three are off unless you turn them on.

What mojibake is

Mojibake is text decoded with the wrong character encoding. The word is Japanese, 文字化け, meaning character transformation, and Japanese text suffers from it often. In English text the usual cause is one mix-up: the text was saved as UTF-8, the encoding nearly everything uses now, and then read as Windows-1252, an older encoding that Windows and many programs still fall back to.

THE MIX-UP

How ’ becomes ’

The characterSaved as UTF-8Read as Windows-1252’E2â80€99™

How ’ becomes ’stringmash.com/mojibake-fixer

Why one character becomes three

In UTF-8 the apostrophe ’ is three bytes, E2 80 99. Windows-1252 has one character for every byte, so it reads those three bytes as three characters: â, € and ™. Every character outside plain ASCII breaks the same way into two, three or four junk characters, and because UTF-8 is so regular, each character always breaks into the same junk. That's what makes it fixable: turn the junk back into bytes and read them as UTF-8.

LOOK IT UP

Common garbled sequences

You seeIt should beCode pointUTF-8 bytes
’’ apostropheU+2019E2 80 99
““ left double quoteU+201CE2 80 9C
â€[U+009D]” right double quoteU+201DE2 80 9D
—— em dashU+2014E2 80 94
–– en dashU+2013E2 80 93
…… ellipsisU+2026E2 80 A6
éé éU+00E9C3 A9
èè èU+00E8C3 A8
áá áU+00E1C3 A1
ññ ñU+00F1C3 B1
üü üU+00FCC3 BC
Â[no-break space](no-break space) non-breaking spaceU+00A0C2 A0
(BOM) byte order markU+FEFFEF BB BF

Common garbled sequencesstringmash.com/mojibake-fixer

Where it comes from

Spreadsheets are the most common source. A CSV file saved as UTF-8 without a byte order mark often opens in Excel as Windows-1252: the Excel CSV guide shows how to open it correctly instead.

Databases are next. A table or connection set to Latin-1 while the application sends UTF-8 stores the bytes faithfully and hands them back mislabelled, and the damage often appears all at once after a migration to a new server.

Then email, where a message's declared character set and its real one can disagree, and copy and paste between programs that make different guesses about a file's encoding.

How to prevent it

Use UTF-8 everywhere and say so. Declare it in web pages with a meta charset tag and in the Content-Type header, set databases and their connections to UTF-8 (in MySQL that's utf8mb4, since its older utf8 can't store emoji), and open CSV files through Excel's import rather than by double-clicking. When a file arrives from elsewhere, check its encoding before you edit and save it, because saving garbled text makes the damage permanent in a new place.

Questions

What is ’ in my text?

A broken apostrophe, ’. The text was saved as UTF-8 and read as Windows-1252. Paste it above to fix it.

Will the fixer change text that's already correct?

No. A repair is only kept when the result is valid and clearly less garbled than before, so café, don’t and naïve pass through unchanged.

Can it fix every kind of garbled text?

It fixes UTF-8 read as Windows-1252 or Latin-1, which covers nearly all garbled English and Western European text. Other mix-ups, such as Japanese or Russian text read with the wrong legacy encoding, need a different repair.

Is my text sent anywhere?

No. The fixer runs in your browser.

Sources

Added . What's new