Repair a corrupted PDF — without uploading

A PDF that will not open is usually a bad day rather than a curiosity — an interrupted download, a failed transfer, a USB stick pulled out early, a file recovered from a dying disk. The document is often irreplaceable, which is exactly why uploading the last surviving copy to an unknown server is a poor way to try to rescue it.

Repair here runs on your device. Ghostscript, compiled to WebAssembly, re-parses whatever it can salvage from the damaged file and writes a fresh, standards-compliant PDF from it — including files other tools reject outright. Nothing is transmitted, because there is no backend on this site to transmit to.

The approach is reconstruction rather than patching. Rather than trying to fix the broken structure in place, the engine reads everything it can still make sense of — page content, fonts, images — and builds a clean document around it. That is why it often succeeds on files that viewers refuse to open at all.

It is not magic, and the honest limits are worth stating:

• Truncated files — if the download stopped partway, everything after the cut-off was never on your disk. Earlier pages usually come back; later ones cannot.
• Destroyed core structures — if the header or the fundamental object table is gone rather than damaged, there is nothing coherent left to rebuild from.
• Encrypted files — these need their password removed first, with Unlock PDF.

You are told how many pages were recovered rather than being handed a file and left to discover the gaps yourself.

Open the repaired file and look through it properly before you delete the original. Recovery is best-effort by nature: pages can come back with missing images, substituted fonts, or content that stops partway down. A file that opens is not automatically a file that is complete.

Keep the damaged original until you have confirmed the repair is genuinely usable. If the first attempt recovers less than you hoped, it is also worth trying to obtain a fresh copy from the source — a re-download frequently beats any repair, since the problem is usually the transfer rather than the document.

Once it opens cleanly, the rest of the toolkit works on it again: split out the pages that survived, merge them with another copy, or extract the text to preserve the content independently of the format.