Mojibake Fixer — Repair Garbled Text Encoding

What it is

The mojibake fixer turns garbled text back into what it was meant to say. It helps when an email subject shows “é” instead of “é”, a database export is full of “’” where apostrophes should be, or Korean and Japanese text has turned into strings of odd symbols. You don’t need to know which encodings were mixed up: every likely pair is tried and the most natural results are listed first.

How to use

1. Paste the garbled text. An example is filled in to start with.
2. Look at the “Most likely” candidate first.
3. Each candidate says how the text was garbled, for example “saved as UTF-8, read as Windows-1252”. If the problem keeps coming back, change that setting in the program that misread it.
4. Copy the candidate you want.
5. If nothing better turns up, or the text is full of �, open the original file with the charset converter.

How it works

• Encodings tried: UTF-8, CP949 (EUC-KR), Shift_JIS and Windows-1252 (Latin-1), all 12 pairs. Each reversal encodes the garbled text with the encoding it was wrongly read as, then decodes the bytes with the encoding they were really in.
• Candidates that lost nothing in the first step are reversed once more to catch double garbling.
• Ranking favours Latin letters and digits, Japanese kana and the 2,350 common Hangul syllables, and penalises �, control codes, stray Latin-1 symbols and Hangul glued in front of Latin letters. A candidate that creates new gaps of � is ranked down hard.
• When 20% or more of the input is �, the “cannot be recovered from the text” note and a link to the charset converter come first, and only candidates whose remaining letters read clearly (common Hangul or kana) are kept, so random-looking noise is never offered as a fix.
• Encoding tables follow the WHATWG Encoding Standard.

Examples

Garbled	Cause	Fixed
Café crème	UTF-8 read as Windows-1252	Café crème
it’s	UTF-8 read as Windows-1252	it’s
한글	UTF-8 read as Windows-1252	한글
À̸§	CP949 read as Windows-1252	이름
譌･譛ｬ隱槭�	UTF-8 Japanese read as Shift_JIS	日本語… (partly)
�ȳ��ϼ���	CP949 read as UTF-8	cannot be recovered (note and converter link shown)