PDF & Image OCR: Text, Tables & Markdown Extractor

Extract text and tables from PDFs, scanned documents and images (PNG, JPEG, TIFF) with Amazon Textract, one of the most accurate OCR engines available. Digital PDFs are read directly from their text layer for $1 per 1,000 pages; only scanned pages and images go through OCR. You get the full text, clean Markdown with tables for LLMs and RAG, and per-page results.

Why this OCR

• Pay for what needs OCR: pages that already contain digital text are read directly at a fraction of the price. Scanned pages in the same PDF are detected and OCRed automatically.
• Tables that stay tables: turn on table extraction to get every table as a cell grid and as Markdown, ready for spreadsheets or AI pipelines.
• Built for documents: invoices, receipts, contracts, forms, reports, research papers and scanned books.
• Up to 500 pages per file and 500 files per run. You can cap pages per file to control cost.

Input

• PDF or image URLs: direct links to PDF, PNG, JPEG or TIFF files, up to 100 MB each.
• Extract tables: layout analysis on every page; tables returned as grids and Markdown.
• Force OCR on every page: for PDFs whose embedded text is broken or garbled.
• Max pages per file: only the first N pages are processed and charged.
• Include per-page output: text and tables for each page separately.

Output example

Real output for the IRS Form W-9 (6 pages, texts shortened):

{
    "url": "https://www.irs.gov/pub/irs-pdf/fw9.pdf",
    "pageCount": 6,
    "pagesProcessed": 6,
    "truncated": false,
    "pagesTextLayer": 6,
    "pagesOcr": 0,
    "tableCount": 0,
    "wordCount": 6279,
    "text": "Form W-9 (Rev. March 2024) Request for Taxpayer Identification Number and Certification ...",
    "markdown": "## Page 1\n\nForm W-9 (Rev. March 2024) ...",
    "pages": [{ "page": 1, "method": "text-layer", "text": "...", "tables": [] }],
    "error": null
}

With Extract tables on, each page includes tables such as [["Item", "Qty", "Price"], ["Widget", "2", "10.00"]], and the Markdown contains them as Markdown tables. Files that can't be downloaded, are password-protected or aren't PDFs or images get an error message and are not charged.

How much does it cost?

Event	Price (free plan)
Run start	$0.02
Page read from the PDF text layer	$0.001
Page or image processed with OCR	$0.004
Page processed with table extraction	$0.03

Paid Apify plans get lower prices automatically. Examples on the free plan:

• A 100-page digital PDF: $0.12
• 100 scanned pages or photos of receipts: $0.42
• 20 pages with table extraction: $0.62

Set Maximum cost per run to cap spending. Each file is only started if the remaining budget covers it.

FAQ

Which OCR engine is used? Amazon Textract, which also reads handwriting reasonably well and handles low-quality scans better than most open-source engines.

Are my documents stored? Files are downloaded to temporary storage only for processing and deleted right after. Results are saved only in your Apify dataset.

Can I use it in AI workflows? Yes. The markdown field is designed for LLM and RAG pipelines. Call the Actor from the Apify API, Make, Zapier, n8n or the Apify MCP server.

What languages are supported? Textract OCR supports English, Spanish, Portuguese, French, German and Italian. The text layer of digital PDFs is read in any language.

Related

From the same author: Audio & Video Transcriber, Amazon Product Scraper, Keyword Research Tool, Website Organic Traffic Checker and AI Search Visibility Tracker.

Support

Open an issue in the Issues tab. Issues are reviewed regularly.