PDF & Scan to Markdown with OCR

Turn PDFs, scanned documents, photos of pages and Word files into clean Markdown or plain text. Give it a list of links; get back one result per document with headings, lists and tables preserved, plus a downloadable .md file. Built for RAG pipelines, AI agents, search indexes and no-code workflows that need readable text instead of a binary file.

No API keys, no browser. Text PDFs are read directly (fast, accurate), and OCR runs only on pages that have no text layer, so you don't pay OCR prices for normal pages.

Why this one?

• Pay per page, and OCR only where needed. $0.003 per page read from a text layer, $0.010 per page that needed OCR. Failed documents and skipped files are free.
• Structure kept. Headings (#, ##), bullet and numbered lists, paragraphs rejoined across line breaks, and tables as Markdown tables.
• Scans and photos. PNG, JPG, TIFF, GIF, BMP and WebP images and scanned PDFs are read with Tesseract OCR in English, German, French, Spanish, Italian, Portuguese and Dutch (combine languages for mixed documents).
• Mixed PDFs handled page by page. A report with a few scanned pages is read from the text layer where it can be, and OCR'd only on the scanned pages. The result shows which method each page used.
• Also reads Word .docx files (headings, lists, tables) and plain text.
• One bad document doesn't stop the run. Every URL gets its own status and a reason if it failed (password-protected, not a PDF, too large, 404...).
• Safe by default. Redirects are checked hop by hop and private networks and cloud metadata addresses are refused.

Input

Field	What it does
Document URLs	Links to the files, one per line
OCR for scanned pages	Auto (default), Always, or Never
OCR languages	eng (default), deu, fra, spa, ita, por, nld; combine for mixed text
Output format	Markdown, plain text, or both
Pages to convert	e.g. 1-5,8,12-. You pay only for converted pages
Include page-by-page output	Adds a pageDetails list with each page's text, method and size
Max pages per document / Max documents / Max file size	Safety limits. Larger files are skipped and not charged
Parallel documents	Default 2 (OCR is CPU heavy)
Extra request headers	Referer, Cookie etc. for sites that check them

Example input:

{
  "documentUrls": ["https://example.com/report.pdf", "https://example.com/scan.png"],
  "ocrMode": "auto",
  "ocrLanguages": ["eng"],
  "outputFormat": "markdown"
}

Output

One dataset row per document:

Field	Meaning
status	converted or failed (with error)
markdown / text	The converted content (documents over 2 MB of text return only the file link)
markdownFile	Link to the .md file in the run's key-value store
pages, pagesTextLayer, pagesOcr	Page counts and how many needed OCR
characters, sizeBytes, fileType	Size and detected type (PDF, image, docx, text)
pageDetails	Optional: per-page text, method and size
pageErrors	Pages that failed OCR, if any

Pricing

Pay per event:

• Page converted: $0.003 (text layer, Word, text)
• OCR page: $0.010 (scanned pages and images)

Word and text files are counted as one page per 3,000 characters. You are never charged for documents that fail.

Good to know

• Tables in text PDFs become Markdown tables when their columns line up. Tables inside scanned pages come back as plain text rows (OCR reads text, not grid lines).
• Heading levels are guessed from text size; a document that uses bold for headings at body size will show them as normal paragraphs.
• Password-protected PDFs and .doc (old Word) files are not supported.
• Handwriting is not reliably read.