Corrupt PDF Repair

What the diagnostic report checks

• Header integrity - whether %PDF- appears where a reader expects it, and whether stray bytes precede it.
• End-of-file marker - whether the file ends with %%EOF as expected, or looks truncated/appended to.
• Cross-reference location and type - whether startxref points at a valid classic xref table or compressed xref stream.
• Object offsets - for classic tables, whether each in-use entry's byte offset actually lands on a matching N G obj header (sampled up to 500 entries).
• Trailer keys - whether /Root and /Size are present so a reader can locate the document catalog.
• Object-count consistency - whether the trailer's declared /Size matches the highest object number actually found in the file.
• Truncated streams - whether the last object's data stream is missing its closing endstream, a common sign of a cut-off download.
• Encryption - whether an /Encrypt dictionary is present, reported for diagnosis only.

Recoverable page preview

Byte-level checks show structural problems, but the preview shows what pdf.js can actually read. The tool opens the file in recovery mode and renders recoverable pages, up to 24 pages. Pages that fail are shown as failed tiles. An encrypted PDF may still have a preview if pdf.js opens it without a password.

How the repair works

For an unencrypted file, the tool attempts to load and re-save it with pdf-lib. This creates a fresh cross-reference table, trailer, and object offsets from the objects the parser can read. Missing or unreadable content is not recreated. The report compares the rebuilt page count with the page count reported by the read-only parser.

When repair is unavailable

Repair is refused whenever an /Encrypt entry is detected. Raw structural checks still run, and page preview may still be available if pdf.js accepts the empty password; otherwise, only the raw checks are available. The tool never removes, bypasses, or guesses passwords.

If parsing or saving fails, no repaired file is offered. The rebuild cannot restore content that is missing or pages that pdf.js cannot open in recovery mode. The diagnostic report remains available to download.

How to repair a PDF

1. Drop the damaged PDF onto the drop zone, or click browse file.
2. Review Diagnostic findings.
3. Inspect Recoverable page preview, including any failed-page tiles.
4. When available, click Download repaired PDF. Click Download diagnostic report for the recorded results.

Privacy & limits

Scanning, previewing, and rebuilding happen locally in browser memory using pdf.js and pdf-lib. The source file is read once and is not modified or overwritten. Full-text structural scanning is capped at 80 MB; the header check uses the first 1,024 bytes and the %%EOF check uses the last 2,048 bytes. Parser-based processing and the repair attempt receive the full byte array. Page preview is capped at 24 pages.

Frequently asked questions