PDF to Markdown & Text with OCR (Tables, Scans)

Turn PDF files into clean Markdown or plain text, ready for ChatGPT, RAG pipelines, search indexes or data extraction. Give it links to PDFs and get back one dataset item per document (or per page) with the text, headings, lists and tables. Scanned pages are read with OCR automatically.

• ✅ Headings, lists and tables become real Markdown (#, -, | … |)
• ✅ Two-column layouts (papers, newsletters, forms) are read column by column, not line across line
• ✅ OCR for scans in English, German, French, Spanish, Italian, Portuguese, Dutch and Polish; mixed PDFs only OCR the pages that need it
• ✅ Clean text: words in tightly set PDFs are split correctly, rotated margin stamps and unmapped glyphs are removed, hyphenated words are rejoined
• ✅ Page ranges (1-3, 7, 10-) and a page limit per document
• ✅ Token estimate per item so you can plan LLM costs
• ✅ Fair billing: you pay per converted page; pages that fail, broken links and non-PDF files are free

Input example

{
    "pdfUrls": ["https://www.irs.gov/pub/irs-pdf/fw9.pdf", "https://arxiv.org/pdf/1810.04805"],
    "outputFormat": "markdown",
    "ocr": "auto",
    "ocrLanguages": ["eng"]
}

Field	What it does
pdfUrls	Direct links to PDF files (up to 100 MB each)
outputFormat	markdown (default) or text
ocr	auto (only pages without a text layer), always or never
ocrLanguages	Languages for OCR, e.g. ["deu", "eng"]
pages	Page range like 1-3, 7, 10-; empty means all pages
maxPages	At most this many pages per document (default 500)
outputPerPage	One dataset item per page instead of per document

Output example

Real output from a test run (the content field is shortened here):

{
    "url": "https://www.irs.gov/pub/irs-pdf/fw9.pdf",
    "fileName": "fw9.pdf",
    "title": "Form W-9 (Rev. March 2024)",
    "pageCount": 6,
    "pagesProcessed": 6,
    "ocrPages": 0,
    "tables": 1,
    "format": "markdown",
    "content": "# W-9\n\n# Request for Taxpayer\n\n… Check only one box on line 3a.\n\n| IF the entity/individual on line 1 is a(n) . . . | THEN check the box for . . . |\n| --- | --- |\n| • Corporation | Corporation. |\n| • Individual or • Sole proprietorship | Individual/sole proprietor. |\n| • Partnership | Partnership. | …",
    "charCount": 37933,
    "wordCount": 6338,
    "approxTokens": 9484
}

A scanned page from the same run (OCR, English):

The LinnSequencer is a state-of-the-art composition and performance tool for the professional musician. It is extremely powerful, yet amazingly simple to learn and use.

A summary per PDF (ok or error with the reason) is saved as OUTPUT in the key-value store.

Pricing

You pay per converted page, plus a small extra per page that needed OCR. In our tests 23 pages from 3 PDFs (one of them scanned) took 12 seconds with the default 1 GB memory.

Paid Apify plans pay less: 10% off on Starter, 20% on Scale and 30% on Business and higher plans.

FAQ

Which tables are detected? Tables with ruling lines (borders or cell lines), like most forms, reports and statistics. Tables drawn without any lines come out as text lines in the right order.

Can it read password-protected PDFs? No. Such files are reported as an error and are not charged.

My scan has a different language. Pick the languages in ocrLanguages; several at once work (["deu", "eng"]), but each extra language makes OCR a little slower.

Can I use files from my computer? Upload them anywhere that gives a direct link (for example an Apify key-value store or a cloud drive share link that downloads the file) and paste that link.

---

Impressum

Levi Griesinger, Tidelane Data
Besigheimer Straße 18
74369, Löchgau, Deutschland
E-Mail: LeviGriesinger@gmail.com