Files are more reliable than pasting. Invisible control characters vanish while text is being pasted, whereas a file is read byte for byte exactly as it was stored.
It did not work
Recovery fails for a small number of reasons, and about half of them are cases where the information was already gone before this tool ever saw it — nothing can bring those back. Find the symptom you are looking at below.
? or � characters in it
Those positions were not turned into a different character — they were thrown away. No information is left, so no tool can restore them. The only way back is to fetch the text again from the original file or screen.
This kind of garbling carries invisible characters, and programs in between silently strip them. The text can look unchanged on screen and still be different by the time you paste it. Copy directly from the screen that holds the original, or use the File tab — files never suffer this damage.
The tool judges how natural a result looks as a combination of characters, and under about ten characters there is not enough to judge from. Paste more text from the same source; longer input is more accurate.
If the lines are broken in different ways they cannot be reversed together. Recovery assumes one single answer to "how was this misread" and rewinds it, so multiple answers cannot hold at once. The tool splits by line automatically, but if that still fails, try one line at a time.
Between single-byte encodings such as Russian, Greek and Hebrew there is almost no signal to tell which one is right. Check the second and third candidates — each one is labelled with the script it was read as.
Nothing is misencoded here — the character information inside the PDF is incomplete. The original is not in the file, so this is out of scope. You would need to find the source document.
?
The characters died when the text was saved. Western European encodings have five empty slots, and about 16% of Korean syllables disappear the instant they land there. Those parts are unrecoverable; only the rest can be salvaged.
So the one reliable fix is to fetch the text again from the original. Almost every case above is "information was shaved off while the text was moved". If you still have the original file, do not paste it — upload it on the File tab. Files are read byte for byte, so the damage that happens during pasting never occurs.
What mojibake is
When text turns into something unreadable like ¾È³çÇϼ¼¿ä or 안, English speakers call it mojibake, borrowed from Japanese 文字化け. Korean speakers have their own name for it — 뷁어 or 뷁체, after the odd syllables that keep showing up in the wreckage.
There is only one cause: the text was stored one way and read another. Computers store characters as numbers, and there are many different tables mapping between the two. Read a file written with an older regional table (EUC-KR, Shift_JIS, GBK) using today's table (UTF-8), or the other way round, and the same numbers come out as different characters. What those tables actually are is explained in UTF-8 vs the old encodings.
Which means nothing was lost — it was just read wrong, and reversing the reading brings the original characters straight back. That is what this tool does. It shows up most often in old forum posts, CSV files written by legacy programs, and filenames inside zip archives.
Files are more reliable than pasting
Text you paste may already have lost something. This kind of garbling contains invisible control characters, and passing through a messenger or a web form silently removes them. Once those bytes are gone there is no way back.
The File tab reads the original bytes directly, so the problem never arises, and one whole guessing step disappears. If you have the file, the file is always the better route. Upload a text file — txt, csv, srt, json, log — and it will be identified and handed back as UTF-8.
If Excel garbles your CSV, Excel is usually the problem rather than the file. It reads unmarked UTF-8 files using the local Windows code page. That is why the download adds a three-byte marker called a BOM — with it, Excel recognises UTF-8. There is a fuller write-up in CSV garbled in Excel, including how to save files that will not break again.
If the names inside a zip archive are garbled, upload the zip itself. The ZIP format has a flag saying how filenames were written, and older compression tools leave it blank, which makes other computers read the names wrongly. This tool fixes the names and repackages the archive. It never decompresses and recompresses, so the file contents are untouched. There is a fuller write-up in garbled filenames inside a zip.
How it works
Garbling has a single cause: data stored with method A was read with method B. So recovery walks the same two steps backwards — the garbled characters are pushed back through the encoding that misread them to restore the original bytes, then read again with the real one. This tool tries every combination of the two and picks the most natural-looking result.
There is no way to guarantee the right answer, so up to three candidates are shown. If the top one looks off, check the second and third.
Common questions
Where does my text get sent?
Nowhere. Pasted text and uploaded files are processed entirely inside your browser, and this site has no server that receives text. Everything disappears when you close the tab and nothing is stored on your device.
How accurate is it?
Across 148 cases in 12 languages, the top candidate is correct 84.5% of the time, and the correct answer is among the three candidates 96.6% of the time. If the top one looks off, check the second and third.
My CSV is garbled in Excel.
Upload it on the File tab and download it with "Add a BOM if you will open this in Excel" left on. Excel reads an unmarked UTF-8 file using the local code page, and those three bytes are what stop it.
The filenames inside my zip are garbled.
Upload the zip file itself on the File tab and it will be repackaged with corrected names. It is never decompressed and recompressed, so the contents are not touched.
Does it work on PDFs?
No. Text copied from a PDF is usually garbled because the character information inside the PDF is incomplete, not because of an encoding mismatch — the original is not in the file, so no tool can bring it back.