PDF Toolkit — OCR, Split, Merge

Three PDF jobs that should never need a subscription: make a scan searchable, pull out a page range, and join files into one. Pay for the pages you actually OCR, or the job you actually run. No monthly fee, no account on a third-party site, no uploading a contract to a free web tool that keeps a copy.

---

What it does

Operation	What you get back
OCR	Your PDF, unchanged to look at, with a searchable, selectable, copy-pasteable text layer underneath. Pages that already have text are left alone.
Split	A new PDF containing only the pages you asked for — 1-3, 1,3,5, or any mix.
Merge	One PDF, files joined in the order you list them.

Output files land in the run's key-value store, ready to download or pass straight into the next step of a workflow.

Input

• Operation — ocr, split or merge.
• PDF file — upload one file, or
• PDF URLs — direct http(s) links, up to 20 files, 64 MiB each. Merge joins them in the order given.
• Page range — split only: 1-3 or 1,3,5.

{
  "operation": "ocr",
  "pdfUrls": ["https://example.com/scanned-contract.pdf"]
}

{
  "operation": "split",
  "pdfUrls": ["https://example.com/report.pdf"],
  "pages": "4-6"
}

{
  "operation": "merge",
  "pdfUrls": [
    "https://example.com/invoice-jan.pdf",
    "https://example.com/invoice-feb.pdf"
  ]
}

Output

One dataset item per result, and the file itself in the key-value store:

{
  "operation": "ocr",
  "sourceUrl": "https://example.com/scanned-contract.pdf",
  "pages": 12,
  "message": "OCR complete (12 pages).",
  "outputs": [{ "key": "ocr-scanned-contract-00-ocr.pdf", "filename": "ocr.pdf", "bytes": 412887 }]
}

An OUTPUT record summarises the run, including anything that was rejected and why.

Pricing

Event	Price
Page made searchable (OCR)	$0.006 per page
Page range extracted (split)	$0.015 per job
PDFs merged	$0.015 per job

A 10-page scanned contract costs $0.06. Pulling three pages out of a 200-page report costs $0.015. You are charged only when the work succeeds — a file we reject costs you nothing beyond the run start.

Limits and guarantees

• 64 MiB (67 MB) per file, 20 files per run, 300 pages overall, 100 pages for OCR (OCR is the expensive one, so it gets the tighter cap).
• Nothing is kept. Each run gets its own isolated working directory, and working files are swept after two hours. Your output lives in your run's storage, under your account.
• Bad input is rejected, not crashed on. A file that is not really a PDF, a corrupt one, an oversized one, or a page range that is not a page range all come back as a plain message saying what was wrong.

Common uses

• Make a folder of scanned invoices or receipts searchable before filing them.
• Split a long statement or report into the section someone actually asked for.
• Merge a set of generated pages into one document at the end of a pipeline.
• Put a text layer on scans so a downstream extraction step has something to read.

Under the hood

ocrmypdf / tesseract for OCR, qpdf for split and merge, poppler for inspection — the mature open-source tools, wrapped so you do not have to install or babysit them. The processing core has been in production use since 7 Sep 2026 and was hardened against a written break-it checklist: non-PDF uploads, corrupt files, oversized files, path-traversal filenames and shell-metacharacter page ranges. subprocess is always called with an argument list, never shell=True.

Not supported yet

Password-protected PDFs, per-page image export, form filling, and languages other than English for OCR. Ask if you need one — they are small additions.