How-to3 min read

File Won't Open: Get Info, Repair, Then Stop Guessing

When a PDF will not open, run Get Info for preflight, then Repair for structural rebuild. That is xref recovery, not a desktop Recovery Toolbox clone.

By the PDF123 team · Updated 2026-10-04
File Won't Open: Get Info, Repair, Then Stop Guessing

A PDF that will not open is usually a structure problem (broken cross-reference table, truncated download, malformed objects), not a missing page image you can redraw by hand. Guessing with OCR, compress, or convert wastes time until the file parses.

Get Info first

Get Info (POST /api/v1/security/get-info-on-pdf) returns a JSON report: FileSize, PageCount, Encrypted, Info dictionary fields when present (title, author, subject, creator, producer), PDF Version, a simple Pages list, and whether a document root was found. Use it as preflight:

  • Encrypted: true means you need an authorized Unlock before most edits; repair alone does not crack passwords.
  • PageCount / FileSize look sane but your desktop viewer still fails: try structural Repair next.
  • The call itself fails to load the bytes: the file may be too damaged for a clean Info pass; still attempt Repair on a copy.

Get Info does not rewrite the PDF. It tells you what the parser can see so you stop treating every failure as "needs OCR" or "needs compress."

Then Repair (structural rebuild via qpdf)

Repair PDF (POST /api/v1/misc/repair) rewrites the file through qpdf so viewers can open it again when enough objects remain. When it works, intended page content is preserved; the job is the map of the file (xref / object streams), not forensic undelete of deleted drafts.

Repair is not:

  • OCR / scan cleanup (file opens but text is only an image): use OCR—which returns Markdown text by default, or a searchable PDF layer if you choose that output
  • Password removal: use Unlock
  • Font repair or "make the file editable again"
  • A clone of the Windows product "Recovery Toolbox for PDF"

For the brand and channel distinction (online structural repair vs desktop "recovery toolbox" searches), read PDF Repair Online vs Recovery Toolbox searches.

Stop-guessing decision tree

  1. Viewer cannot open the file: Get Info, then Repair.
  2. File opens but select/search fails on a scan: OCR for Markdown text or a searchable PDF; optionally Compress a copy if size is the remaining problem (Scan pack: OCR then Compress).
  3. File opens, tables export as 204: no columns detected; see Empty table export (204).
  4. Same repair from a script or agent: POST /api/v1/misc/repair via Developers. Send an Idempotency-Key if your client retries uploads (Idempotency-Key).

What “repaired” still will not fix

A successful Repair means a viewer can open the file again. It does not mean:

  • Every page image was reconstructed from nothing (missing objects stay missing)
  • Text became selectable on a scan (that is OCR, which returns Markdown or a searchable PDF here)
  • Encryption disappeared (Unlock with a known password)
  • Tables will export (still need columnar text; 204 means none detected)

If Get Info shows a sane page count after Repair but a page is blank, the structure recovered and that page’s content may still be empty or truncated in the original bytes.

Keep the damaged original

Keep the damaged original beside any repaired derivative. If Repair cannot recover enough structure, the honest outcome is "bytes are gone," not another random tool in the catalog. Compressing or converting a file that still will not parse does not invent missing objects.

Compare channel intent (online structural repair vs desktop “recovery toolbox” brand searches) in PDF Repair Online vs Recovery Toolbox searches if that is how you found the tool. No account is required for the browser path on catalog tools.

Open tool
Process in the browser — no watermark, files removed after the job.
Open tool