PDF Text & Table Extractor - OCR, Markdown for RAG

Turn PDF files into clean text, tables and Markdown. Paste direct PDF links, get structured data back: one row per page or one row per document. Scanned pages are recognised automatically with OCR in 25+ languages. Built for RAG pipelines, document search, data entry automation and research.

What you get

• Text of every page, with the original layout preserved (columns stay aligned).
• Tables as arrays of rows, detected from the page layout, ready for CSV/Excel or a database.
• Markdown with joined paragraphs, headings and Markdown tables - ready to chunk for LLMs and vector search.
• OCR for scanned pages (Tesseract): English, German, French, Spanish, Italian, Portuguese, Russian, Ukrainian, Polish, Dutch, Turkish, Arabic, Chinese, Japanese, Korean, Hindi and more.
• Metadata: title, author, creation date, producer, page count, file size.
• Page ranges such as 1-5 or 3,7,10- and a cap on pages per file.
• Bulk mode: hundreds of PDFs in one run, processed in parallel.

How it works

1. You pass direct links to PDF files.
2. The actor downloads each file and reads its text layer page by page.
3. Pages without a text layer (scans, photos) are rendered and recognised with OCR.
4. You receive the results in the dataset (JSON, CSV, Excel, XML, HTML) and via API.

Input

Field	Description
PDF URLs	Direct links to PDF files (required).
Page range	For example 1-5 or 1,3,10-20. Empty means all pages.
Output format	One row per page, or one row per document.
Detect tables / Return Markdown	Turn table detection and Markdown on or off.
OCR for scanned pages	On by default. Pages with almost no text layer are recognised with OCR.
Force OCR	Recognise every page, ignoring the embedded text.
OCR languages	For example eng or eng+deu.

Example input:

{
    "startUrls": [{ "url": "https://arxiv.org/pdf/1706.03762" }],
    "pageRange": "1-3",
    "outputMode": "perPage"
}

Output (one row per page)

{
    "url": "https://arxiv.org/pdf/1706.03762",
    "fileName": "1706.03762",
    "title": "Attention Is All You Need",
    "pageCount": 15,
    "page": 1,
    "ocrUsed": false,
    "charCount": 4120,
    "text": "...",
    "markdown": "...",
    "tables": [],
    "scrapedAt": "2026-10-14T08:00:00.000Z"
}

In the per-document mode you get one record with the whole text, the Markdown, all tables with their page numbers and a list of pages.

Pricing

You pay only for pages that returned text. Pages with a text layer and OCR pages have separate prices; empty pages and failed files are free.

Event	Price
PDF page with text layer	$0.004 per page
OCR page (scanned)	$0.008 per page

Volume discounts apply automatically for Bronze, Silver and Gold Apify subscribers. A 20-page report costs about $0.08. Set a maximum cost per run in the run options to keep full control.

Tips

• The link must return the PDF file itself. If a site shows a viewer page, copy the direct .pdf link.
• For scans in a language other than English, set the OCR languages, for example deu or eng+fra.
• For RAG, use the per-page output and chunk the Markdown field.
• Password-protected PDFs are not supported and are reported as errors.

Limits and notes

• Table detection reads the text layout. It works well on reports, invoices and statements with text columns; borderless or heavily merged tables may need a manual check.
• OCR quality depends on the scan. Use 300 DPI for small print.
• Only publicly reachable files can be processed. Please make sure you have the right to process the documents you submit.

Integrations

Run it from the Apify API, schedule it, or connect it to Make, Zapier, n8n, Google Sheets or your own code with the Apify client for JavaScript and Python.

FAQ

Does it work with scanned PDFs? Yes. Pages without text are recognised with OCR automatically.

Can I get tables as CSV? Yes. Export the dataset as JSON or CSV; tables are returned as arrays of rows and can be converted in one line.

What is the difference between page and document output? Page output suits RAG and search; document output suits archiving and one-record-per-file workflows.

Something looks wrong on my file. Open an issue in the Issues tab with the PDF link and we will look at it.

More tools by the same author

• Image OCR Bulk: Bulk OCR for images
• Markdown & HTML to PDF: Print-ready PDF from Markdown and HTML
• Bulk Image and File Downloader: Bulk download of images and files
• Website Screenshot API: Full-page and mobile screenshots in bulk