Garbled characters are turned back into the bytes they came from using the table of the encoding that was wrongly used, then read again strictly with the right one. Only results whose bytes reproduce your input exactly become candidates.
Undo mojibake,
from the original bytes.
Diagnose text garbled by a UTF-8 / Shift_JIS (CP932) / EUC-JP / UTF-16 mix-up and see only the candidates that truly reverse it, with the evidence. CSV, TSV and TXT files can be converted too.
Parts that became 「�」 or Excel’s 「・」 no longer hold the original data. They are shown as unrecoverable instead of being filled with plausible characters.
Text and files are processed on this device. No server upload, no sign-up and no AI.
How to use it
- Paste the garbled text, or open a CSV, TSV or TXT file in the “Open a file” tab.
- Diagnose it: only candidates whose reversal matches byte for byte are shown.
- Pick a candidate and check the comparison with your text and the evidence. Lost parts stay as 「�」.
- Check the content, then copy the result or save it.
FAQ
Which kinds of mojibake are supported?
Text garbled by a UTF-8 / Shift_JIS (CP932) / EUC-JP / UTF-16 mix-up (such as 「縺ゅ」 or 「ã‚」), including the same mix-up applied twice. For other causes the tool says that no pattern was found.
Is a candidate guaranteed to be the original text?
No. The tool checks that reading the candidate’s bytes with the wrong encoding reproduces your input character for character (an exact reversal). It can’t prove that this is the original text, so check the content before using it.
Can characters that became 「�」 or 「?」 be recovered?
No: the original bytes were lost when they were replaced. They are shown as lost instead of being guessed. If you still have the original file, opening it as a file may read it correctly.
Are my files uploaded?
No. Text and files are processed only in your browser and never sent to a server.