Repair PDF

How to repair a PDF that will not open

Drop the file in. It is examined before anything is changed, and what comes back is a list rather than a verdict: does it start the way a PDF must start, is the end still there, how many objects and pages can be found inside it, and will an ordinary reader open it. Then you are told what can be done, and only then is there a button.

1. Drop the damaged PDF on the box above. Nothing is sent anywhere.
2. Read the list of what is wrong. Each line is either a tick or a cross with the reason beside it.
3. Press Repair it. The file is trimmed, read forgivingly, and written out again with a fresh index.
4. If the document cannot be rebuilt but its pages can still be drawn, press Rescue the pages as pictures instead.
5. Open the saved file before you throw the old one away.

What actually breaks in a PDF

Almost every damaged PDF is damaged in one of four dull ways, and none of them means the pages are gone.

Something is stuck to the front. A PDF has to begin with the five characters %PDF-, and a file that has been through a badly written server or mail program can arrive with headers, a stray blank line or a page of HTML in front of them. Strict readers refuse it on sight. Cutting the rubbish off fixes it completely.

The end is missing. The index that says where every object in the file lives sits at the end of a PDF, not the beginning - so a download that stopped early, or a file copied off a failing disk, loses the map rather than the contents. Everything is still in there; nothing can find it.

The index is wrong. It says object 12 lives at byte 90,000 and it does not, usually because something rewrote part of the file without rewriting the index to match.

One object inside is malformed, and a strict reader stops dead at it rather than stepping over it and carrying on.

The repair is the same in all four cases: trim the file back to the PDF inside it, read it with every allowance the library can be given, then copy the pages into a brand new document. Copying pulls across only what each page actually points at, so anything orphaned or broken beyond understanding is left behind rather than carried into the new file, and the index is built fresh.

When the pages can be rescued but the document cannot

Sometimes a file is too far gone to be rebuilt as a document, but a reader can still draw the pages on screen. In that case there is a second button, and it does something cruder: every page is drawn and photographed into a new PDF.

You get your pages back, and you lose the words. They stop being text and become part of a picture, so they can no longer be searched, selected or copied. It is worth knowing that the OCR tool on this site can read them back afterwards, which puts most of what was lost back into the file.

It is offered second, and described plainly, because it is a worse outcome than a real repair and nobody should end up with it by accident.

What this cannot do

If neither reader here can find anything in the file, you will be told so rather than handed a button that appears to work. A file that has been truncated to a fraction of its size, or overwritten by something else, or filled with zeroes by a dying disk, does not have a document in it any more, and no repair tool of any price can invent one.

When that happens, any other copy you have is worth more than anything this page can do: the original email attachment, a backup, a cloud folder, the downloads folder on the machine that made it, or the program that saved it in the first place, which may still have its own recovery file.

A password-protected file can be repaired, but the result is still password-protected. This site does not remove passwords.