How to use
- Drop the ZIP that will not open, or click to choose it. Self-extracting (.exe) files are accepted too.
- Check the Diagnosis table: end record (EOCD), central directory, truncation, bytes before the ZIP and split-archive status.
- In the results table every file shows its status (Good ✅ / CRC mismatch ⚠️ / Unrecoverable ❌). Tick what you want and press Download selected, or press Build repaired ZIP to pack only the good files into a new ZIP.
One selected file downloads as is; several come as a new ZIP with UTF-8 names.
How the repair works
A ZIP stores a local file header (PK\x03\x04, a 30-byte fixed part plus the name) in front of every file, and a central directory (the list) plus an end record (EOCD, 22 bytes) at the very end. Most programs read the list first, so when the end is cut off or the list is damaged they say the archive cannot be opened, yet the data of each file often sits intact near the start.
| Stage | When | Method |
|---|---|---|
| 1 | The central directory can be read | Extract each entry from the list and compare the CRC-32 |
| 2 | The list is missing or damaged | Search the whole file for local headers and recover from those |
In stage 2, if a header holds the compressed size, that many bytes are read. Entries that use a data descriptor (bit 3) have size 0 in the header, so the candidates for their end are the next header, the central directory or the end of the file; the end is accepted where the 12- or 16-byte descriptor just before it states a compressed size that equals the real distance. That is why a stored entry containing a whole other ZIP does not make the inner entries look like outer ones. The descriptor’s CRC is then compared with the CRC of the extracted data.
Reading the status
| Status | Meaning |
|---|---|
| Good | Size and CRC-32 both match. If no CRC could be found (for example for a last entry) but the compressed stream decoded completely, “(CRC not available)” is added |
| CRC mismatch | Data came out but part of it changed. Photos and documents may look partly broken. It goes into the repaired ZIP only if you switch the option on |
| Unrecoverable | The file is cut off or its compressed data is damaged |
Junk in front of the ZIP is handled too. If 1,024 bytes of other data precede the archive, the diagnosis shows “Bytes before the ZIP: 1,024” and reading starts from there. Zero padding after the end record is fine; non-zero junk after it makes the tool distrust the end record and recover by scanning headers instead.
What cannot be saved
- Entries cut off at the end: if compressed data stops halfway, that entry is lost. Intact entries before it are kept.
- Split archives (.z01, .z02 … .zip): join the parts into one file first. From a single part only some files can be saved.
- Password ZIPs: they appear as “Encrypted entry” and open once you type the password. ZipCrypto and AES are supported and the password is used only on this device.
- Unsupported compression methods (BZip2, Deflate64 and so on) mark only that entry as unrecoverable.
Safety measures
Recovered content lives only in memory and is never written to disk. ../, absolute paths and drive letters are removed from names to stop path escapes (zip slip), and duplicate names become a (2).txt. Output is capped at 1 GB in total; an entry that expands to at least 1 MB at a ratio above 1000x may be a zip bomb and is skipped (you can include it after confirming with “Extract anyway”). A header that lies about the size cannot make the tool inflate beyond the size it declared.
To make a clean ZIP use ZIP Creator, to open a healthy archive ZIP Extractor, to check what a file really is File Type Checker, and for a broken PDF Repair PDF.
Files are processed only in your browser and are never sent to a server.