Repair PDF

How it works

Open the damaged file. The tool examines it and explains in plain words what is wrong: a broken cross-reference table, a cut-off file, junk at the start or end, wrong data lengths, a damaged page tree or pages with unreadable content. If the file is healthy, it says so and changes nothing. If the file will not open at all, the page says so and offers “Try to repair it”: the same rebuild is then made straight from the bytes of the file, and if nothing can be recovered it says plainly what was tried.

When you repair it, every object still present in the file is found, the structure is rebuilt and a completely new file is written. The result is then reopened with our own readers and the report lists the pages recovered out of the estimated total, the objects dropped, the text that can be extracted and the signatures that are no longer valid. The original is not touched: the result is a new version.

If rebuilding is not possible you can ask, as a last resort and only by your own choice, for a lossy recovery: each page is drawn and a PDF made of pictures is created. It loses the text, links, forms and bookmarks, and the tool says so clearly. It never runs by itself.

What is not preserved, and limits

• Content that is physically gone from the file cannot be recovered. Only the readable part of a damaged data block is kept.
• A block cut in the middle loses what came after the cut, and the page that used it may come out blank or incomplete. A cut picture shows only what survived.
• When the page tree has to be rebuilt, the pages follow the order of the file, which is not always the original order.
• Objects deleted by an earlier edit can come back, because the newest copy of every object found in the file is used.
• Fonts and images that the pages refer to but that are missing are not recreated: the text may show in a substitute font or blank.
• Digital signatures stop being valid after any rewrite, and a certified document is no longer certified. The tool warns you before it starts.
• A password-protected file needs its password and comes back without it. If the end of the file that describes the encryption is also lost, it cannot be repaired.
• The result is checked with our own readers, written by the same team: that does not prove it opens in every program. Always open it in the viewer you use.
• A page that reads correctly but holds wrong instructions or altered letters inside its text cannot be told apart from a good one. The tool cannot guess the original content.
• The lossy recovery as pictures uses another reader that sometimes also refuses the file (for example, a truncated one): the pages it cannot read are then left blank and listed.
• Tested with deliberately damaged files judged with pikepdf, qpdf, poppler and Ghostscript; it has not been tested with Adobe Acrobat or Reader. Everything happens in your device's memory and nothing leaves it; the limit is about 600 MB per file.