How to use
- Drop a PDF that will not open (or click to choose it). As soon as it loads, a diagnosis list appears.
- Click Try to repair. The steps below run automatically, and later steps are skipped once one works.
- On success you get
name_repaired.pdf, a table of what worked and failed, and the page count (estimated before → after). - If every step fails, you see “Could not repair this file” with likely causes.
Repair steps
| Step | What it does | Text and images |
|---|---|---|
| 1. Diagnose | Checks the header position, the end of file (%%EOF), the startxref value, and whether it is HTML or ZIP | Unchanged |
| 2. Rewrite structure | Reads the file with a forgiving parser and saves a fresh structure; junk at the start is cut off | Kept |
| 3. Rebuild xref | Scans the whole file for N G obj … endobj, builds a new cross-reference table and trailer, then repeats step 2 |
Kept |
| 4. Pages as images | Renders only the pages the browser’s PDF engine can open into a new PDF | Text cannot be selected or searched |
Each step counts as a success only after the saved file is opened again and its page count is checked. Pages that cannot be rendered in step 4 are left blank and listed as “pages N could not be recovered”. You can choose step 4’s resolution (150, 200 or 300 dpi, default 200) and JPEG quality (default 85%).
Common causes of damaged PDFs
- Interrupted download: the end of the file is missing, so there is no
%%EOFor page location table. - Damaged email attachment or chat transfer: extra bytes are added at the start or line endings are converted, so stored positions no longer match.
- Closed while saving: the program quit mid-save and only half of the file was written.
- Saved web page: a login or error page is downloaded with a
.pdfname (shown as HTML in the diagnosis).
Example results
These are results from cutting a simple 5-page test PDF created with this tool (no object streams):
| Portion kept | Diagnosis | Result |
|---|---|---|
| 100% (intact) | No problem | 5 pages from the structure rewrite |
| 70% | End of file cut off | 4 pages from the structure rewrite |
| 40% | End of file cut off | 1 page from the structure rewrite |
| 10% | End of file cut off | Cannot be repaired (no openable page) |
Real files differ with their structure. This table shows values for a simple test file; in general, pages after the cut-off point cannot be brought back. If the page count dropped, download the original again and compare.
When repair is not possible
- Files that barely downloaded (like the 10% case above, with no page information).
- Files that are not PDFs, such as HTML, ZIP or empty files: the diagnosis says so and no repair is attempted.
- Password-protected files: the structure can be fixed without the password, but reading page content, as in step 4, needs it. The original password still applies after a structure repair.
- Files whose compressed data inside is itself damaged: the structure can be fixed while that page’s content may still look empty.
Good to know
- A repaired file can differ in size from the original. To shrink it, use Compress PDF.
- To remove a password you know or lift editing limits, see Unlock PDF. To get pages as pictures, use PDF to JPG.
- Files up to 300 MB and 2,000 pages are supported.
Your file is processed only in your browser and is never sent to a server, so sensitive documents are safe to repair here.