01
Prefer the original file when it survives
File and Hex mode retain real bytes and report zero-based UTF-8 error offsets. A text paste only contains the characters your browser received, so it cannot provide historical byte positions.
Unicode and Internationalization
mojibake fixertrace the wrong encoding
Compare repair paths against the untouched text or original bytes.
mojibake fixer
Garbled text or original bytes
Repair candidates
Enter text, paste Hex, or choose a file.
Text and file bytes are processed in this browser; nothing is uploaded.
Input or settings changed. Waiting for a fresh analysis.
Candidates are guesses. Reversibility does not prove which reading is historically correct.
A mojibake fixer helps when a name becomes café after a file is opened with the wrong character set. Start with the smallest broken passage, then compare candidates against words you know should be present.
If you still have the file, use its bytes instead of a pasted replacement string. That lets you fix garbled text by reinterpreting the source, while keeping a copy of the exact bytes that caused the problem.
Why use this mojibake fixer?
Keep the evidence visible while trying a different interpretation.
One byte sequence, several readings
The mojibake fixer keeps a local file or Hex sequence intact while UTF-8, GBK and other decoders produce separate previews. A readable preview is a hypothesis; the original bytes remain available for a second look.
Reverse the mistaken reading
For pasted café, the tool maps the visible characters back through the selected mistaken encoding and then decodes those bytes as the intended encoding. It shows the assumed chain instead of silently replacing your input.
Choose evidence before a repair path
02
Read the direction labels literally
Automatic mode tries reversible paths. In manual mode, Written as names the original encoding and Read as names the mistaken interpretation; changing either setting refreshes candidates after a short pause.
03
Compare words, then export one result
A reversible chain only reproduces the current string when run backward. Check names and punctuation yourself; copy or download is disabled whenever the input or options make the previous result stale.
Evidence and candidates in one place
Strict UTF-8 validator
For a file or Hex input, invalid lead, continuation, truncated and out-of-range bytes are reported by offset. This helps fix broken UTF-8 without pretending a pasted string still contains those bytes.
Several repair directions
The candidate list covers Windows-1252, Latin-1, Cyrillic and common East Asian readings. It can expose double-encoded text without automatically choosing a supposedly certain original.
Local byte preservation
Switch decoders without altering the source array. Download the original byte copy separately from the selected UTF-8 text, so a failed guess does not erase evidence.
What the result can and cannot prove
- A candidate is not a verdict
- Even a reversible candidate may be the wrong language or encoding. Short strings can look plausible under several decoders; use surrounding words or the original file's metadata to decide.
- Replacement characters mean lost information
- If another program already inserted � or ?, the original byte value is not encoded in that character. The tool flags � and cannot promise a full recovery from pasted text alone.
- Byte offsets require bytes
- UTF-8 error offsets are computed only for file or Hex input. A Unicode text paste has no access to the source file's invalid byte positions.
- No file overwrite or legacy export
- The page creates a new UTF-8 text download and, when available, a copy of original bytes. It does not rewrite a source file, database row, PDF, or legacy encoded file.
Mojibake fixer questions
Why did my text get corrupted?
Text is often corrupted when bytes written in one encoding are decoded with another. Try the original file bytes when possible; a pasted string may have passed through several conversions already.
Is the repair always possible?
No. A missing byte cannot be inferred reliably after it becomes � or ?. Compare candidates with context, and return to the source export if the needed bytes are gone.
Will it damage my correct text?
Analysis never rewrites input. Choosing a candidate changes only the preview; copying or downloading creates a separate result, and you can also save the untouched original bytes.
Only part of my document is broken. What happens?
Work on a small broken excerpt first. The page does not find and rewrite mixed encodings across an entire document or database; a long file may contain distinct regions.
There are still odd characters in the result. Why?
The selected chain may be wrong, the text may have suffered more than one conversion, or a replacement character may already have erased data. Review another candidate and the original bytes.
Still seeing a broken fragment?
Try the mojibake fixer on your fragment
Use this mojibake fixer to compare the likely readings, then save only the candidate that matches your real context.