Repair a damaged PDF

Fix a PDF that won't open. Rebuilds a damaged file structure automatically, and recovers unreadable pages as images with OCR text when needed. Runs on our server; files are deleted automatically within about two hours.

Drop your files here

.pdf · up to 300 MB · up to 30 jobs an hour · uploaded to our server and deleted after processing; the result is wiped within about two hours

Choose files

How to Repair the PDF

1
Upload the damaged PDF
A file that won't open, opens with errors or shows garbled pages, up to 300 MB.
2
Choose a repair strategy
Auto tries a full structural fix first and only turns damaged pages into scanned images if it has to. Two other modes give you more control either way.
3
Download the fixed file
The result screen says what was repaired: a clean structural fix, or which pages had to be recovered as images.

It doesn't fix everything, and says so

Most real-world PDF damage is a broken cross-reference table or trailer, usually from an interrupted download or a failing disk. That is the case this tool fixes completely, restoring the file to its original content. A damaged content stream on a page is different: the vector text and graphics on that page are gone, and no tool can reconstruct data that no longer exists. When that happens, the affected pages are rasterized into images and passed through OCR to recover readable text, while every unaffected page keeps its original vector content untouched. The result screen says which of these happened, page by page.

Three strategies, three trade-offs

Auto is the default: it rebuilds the file structure first and falls back to image recovery only on the pages where the content stream itself is damaged. Everything else stays as it was. Structure-only skips image recovery entirely; pages with a damaged stream come back blank instead, which is the right choice when you need a clean text layer and would rather know a page is missing than get a scanned image of it. Rasterize-and-OCR is the opposite: every page becomes an image with an OCR text layer, even pages that were readable. That helps when the visible text looks like garbled characters rather than an outright rendering error, which structural repair alone will not catch.

Processing happens on a server, not in your browser

Parsing malformed PDF byte structure and partially decoding broken content streams is something no browser-based PDF library does as reliably as a dedicated server binary, so this tool is one of a small number on PulsarKit that uploads your file for processing rather than working entirely on your device. The uploaded file is deleted as soon as the job finishes, and the repaired copy is wiped automatically within about two hours.

Processed on our server, deleted automatically

The PDF is uploaded only for the repair itself. The uploaded file is deleted as soon as the job finishes. The result stays downloadable for one hour and is wiped from the server within about two hours, whether or not you download it. There's no account and no manual cleanup step required.

Frequently asked questions

Why doesn't this tool fix every broken PDF?

A damaged content stream means the page content itself, the text positions and drawing instructions, is corrupted or gone. That is not a structural problem to repair; it is missing data. The best recovery available is to turn what can still be rendered into an image and add OCR text on top, which is what this tool does automatically when needed.

What's the difference between Auto, Structure-only and Rasterize-and-OCR?

Auto rebuilds the structure and rasterizes only the pages that are damaged. Structure-only never rasterizes; damaged pages come back blank instead, for when you need a clean text-only result. Rasterize-and-OCR converts every page to an image with OCR text, which helps when text renders as garbled characters rather than an outright error.

What does it mean if a page became an image?

That page's original content stream was damaged beyond structural repair. The tool rendered what could still be recovered from it, added a text layer with OCR, and tells you which pages this happened to.

Can I get the original vector text of a damaged page back?

No. If a page's content stream is damaged, that data is gone. Rasterizing with OCR is the most complete recovery available, but it is not the same as the undamaged original.

How long is my file kept on the server?

The uploaded PDF is deleted as soon as the job finishes. The repaired file stays downloadable for one hour and is wiped from the server within about two hours, whether or not you download it. Nothing is kept longer than that.

What if the file can't be recovered at all?

If not even one page can be recovered, even as an image, the tool tells you so rather than returning a broken or empty file. At that point the file itself is likely truncated or overwritten, and restoring from a backup is the only path back.

Is it safe to upload a PDF with confidential information?

The file is processed solely to repair it. The upload is deleted as soon as the job finishes and the repaired copy is wiped within about two hours; it isn't used for anything else and isn't kept beyond that window.