Repair PDF
Rebuild a damaged file so it opens again.
This tool runs entirely in your browser. Your file is never uploaded, and you can verify that in your browser's network tab. Verify it yourself: open your browser's network tab and watch. You will see one small request asking whether you have tasks left - a tool name and a hash, never the file.
What this does
This rebuilds a damaged PDF from whatever is still readable inside it. The cross-reference table - the index that tells a reader where every object lives - is reconstructed by scanning the file, and the result is written out as a clean document. Before you are offered anything, the repaired file is opened again to check it really works, and the report tells you how many pages it had before and how many it has now.
Use it when a reader refuses the file: a download that stopped part way, a file recovered from a failing drive or a memory card, one that came out of a crashed export, or one that opens in a browser but not in Acrobat.
How it works
- Drop the damaged PDF onto this page. This is the one tool that accepts a file whose header looks wrong, because that is exactly the file it exists for.
- Leave Rebuild the cross-reference table on. It is the fix for the large majority of PDFs that will not open.
- Leave Recover pages on so that, if the ordinary repair fails, the document can be rebuilt from its pages alone.
- Turn on Drop unreadable objects only if the first attempt fails. It sacrifices whatever those objects held in exchange for getting a file at all.
- Press Repair, then read the report: it says how many pages were readable before and how many the repaired file contains.
Most files that will not open are not badly damaged. A PDF ends with a table of byte offsets pointing at every object in it, and if that table is wrong - because bytes were added, a transfer truncated, or an editor wrote it incorrectly - a strict reader gives up even though the content is all present. The repair scans the whole file for objects, rebuilds the table from what it finds, and writes a fresh document around it.
There is a second, stronger attempt behind the first. If the ordinary repair cannot complete, the document is rebuilt from its page tree into an empty file, which recovers documents whose catalogue is gone but whose pages survived - the usual shape of an interrupted download. That path copies page content and leaves document-level structures behind, so a file recovered this way may lose its outline, its interactive form and its attachments while keeping every page.
The result is verified before it is offered to you. A repair that produces a file nothing can open has not repaired anything, so the output is read back and its pages counted. The counts before and after are both reported, and a repair that recovered 4 pages of 40 is a very different outcome from one that recovered all 40.
Nothing can be recovered that is not in the file. If a transfer stopped at 60 per cent, the missing 40 per cent is gone and no tool will invent it. Keep the damaged original until you have checked the result, because a second attempt with different options sometimes does better.
The repair runs on your device using qpdf compiled to WebAssembly. That matters more here than elsewhere: a file you cannot open is a file you cannot check, and uploading an unknown document to a stranger's server to find out what is in it is a poor trade.
What this tool cannot do
- Data that is genuinely missing cannot be recovered. A truncated download comes back as the pages that arrived, not as the whole document.
- An encrypted file cannot be repaired without its password, because its objects cannot be read. Remove the protection with Unlock PDF first.
Questions people ask
- What is actually wrong with a PDF that will not open?
- Usually the cross-reference table, which is the index of byte offsets at the end of the file. If it disagrees with where the objects really are, a strict reader refuses the document even though the content is intact. Rebuilding that table by scanning the file fixes the majority of these cases.
- How many of my pages will come back?
- As many as are still readable, and the result tells you the number. It reports the pages that could be read before the repair - often zero, since the file would not open - and the pages the repaired document contains. If those numbers do not match what you expect, part of the file is missing rather than merely broken.
- The repair failed. What else can I try?
- Turn on Drop unreadable objects and run it again: objects that cannot be parsed are discarded instead of stopping the job, which loses whatever they held but often produces a usable document. If that still fails, compare the file size against the original source - a file much smaller than it should be was truncated in transfer, and re-downloading is the only fix.
- Do I have to upload the broken file?
- No, and that is worth something here. A file you cannot open is a file you cannot inspect, so sending it to an unknown server to find out what is inside it is a risk you do not need to take. qpdf runs as WebAssembly in your browser and nothing is transmitted.
- Does repairing change the pages themselves?
- No. Page content is copied as it is; what gets rebuilt is the structure around it. If the stronger recovery path was needed, document-level features such as the outline, the interactive form and any attachments may not survive, but the pages come across unchanged.