PDF Text Extractor with OCR

Extract clean text from PDF files: paste PDF links or upload files and get the text of every page back as JSON, CSV or XLSX. Normal PDFs are read directly from their text layer, which is fast and cheap. Scanned pages are detected automatically and read with OCR. You only pay for OCR on the pages that need it.

Built for: RAG and AI pipelines, document search, contract and invoice processing, research archives, and anyone who needs PDF text without copy-pasting.

What you get

• Smart OCR. Each page is checked. Pages with real text are read directly. Scanned or image-only pages, and pages with broken font encoding, go through OCR (Tesseract).
• Mixed PDFs handled. A PDF with typed pages and scanned pages is read correctly page by page.
• One result per PDF or one per page. Per-page output drops straight into embeddings, vector databases and spreadsheets.
• Page numbers, method and confidence. Every page says whether it was read from text or OCR, with an OCR confidence score.
• 7 OCR languages: English, Spanish, French, German, Portuguese, Italian and Dutch, mixable in one run.
• Page ranges and limits. Read pages 5–20 only, cap pages per PDF, cap file size.
• PDF metadata: title, author, creator, producer and dates when present.
• Fair billing. Blank pages, failed pages and files that can't be opened are never charged. The run stops cleanly at your maximum cost per run.

How to use

1. Add PDF links (one per line) or upload PDF files in PDF files.
2. Leave OCR for scanned pages on Auto, or choose Never to skip OCR entirely.
3. Pick your OCR languages if documents aren't in English.
4. Choose One result per PDF or One result per page.
5. Click Start, then download results as JSON, CSV, XLSX or HTML, or fetch them with the API.

Use it from code

curl -X POST "https://api.apify.com/v2/acts/ridgeline_apps~pdf-text-extractor/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"files": ["https://example.com/report.pdf"], "outputMode": "page"}'

It also works from Make, Zapier, n8n and the Apify MCP server for AI agents.

Input

Field	Description	Default
files	PDF URLs or uploaded files	required
ocrMode	auto, never or always	auto
ocrLanguages	Any of eng, spa, fra, deu, por, ita, nld	["eng"]
outputMode	document (one result per PDF) or page (one per page)	document
includePageDetails	Per-page list inside each document result	true
pageFrom / pageTo	Page range to read	all pages
maxPagesPerPdf	Safety cap on pages per PDF	500
maxFileSizeMb	Larger files are skipped (free)	50
ocrDpi	OCR resolution, 150–400	300

Output

One result per PDF (shortened):

{
  "url": "https://example.com/contract.pdf",
  "fileName": "contract.pdf",
  "status": "ok",
  "pageCount": 3,
  "pagesProcessed": 3,
  "textPages": 1,
  "ocrPages": 1,
  "emptyPages": 1,
  "wordCount": 412,
  "text": "MASTER SERVICES AGREEMENT\nThis agreement is made between...",
  "metadata": { "title": "Contract", "producer": "Example Writer 2.1" },
  "truncated": false,
  "pages": [
    { "pageNumber": 1, "method": "text", "text": "MASTER SERVICES AGREEMENT...", "ocrConfidence": null },
    { "pageNumber": 2, "method": "ocr", "text": "Signed by both parties...", "ocrConfidence": 94.1 },
    { "pageNumber": 3, "method": "empty", "text": "", "ocrConfidence": null }
  ]
}

One result per page:

{ "url": "https://example.com/contract.pdf", "fileName": "contract.pdf", "pageNumber": 2, "pageCount": 3, "method": "ocr", "text": "Signed by both parties...", "wordCount": 210, "ocrConfidence": 94.1, "status": "ok" }

Files that fail come back with "status": "error" and a plain-English error such as "This PDF is password-protected." They are never charged.

Pricing

Pay per page, with no subscription:

• Text page (read from the PDF's own text): charged per page.
• OCR page (scanned page read with OCR): charged per page.
• Blank pages, failed pages and unreadable files: free.

The exact prices are shown on the Pricing tab. A 10-page typed PDF costs a fraction of a cent, and you can set a maximum cost per run in the run options.

FAQ

Does it work on scanned PDFs? Yes. With OCR on Auto, image-only pages are detected and read with OCR. OCR accuracy depends on scan quality; clean 300 DPI scans usually score above 90% confidence.

Does it keep tables and columns? Text keeps its line breaks and reading order as stored in the PDF. Tables come out as text lines, not as structured cells.

Password-protected PDFs? They're reported as errors and not charged.

Handwriting? OCR is designed for printed text. Handwriting results are poor.

Where do my files go? Files are downloaded into your own Apify run and processed there. Nothing is sent to any third-party service. Results stay in your Apify storage under your account's retention settings.

Limits? Up to 200 MB per file and 5,000 pages per PDF. For large batches, give the run more memory for faster OCR.

Legal

Only process documents you have the right to use. You're responsible for the documents you submit and how you use the extracted text. Not intended for medical records (PHI) or other regulated data that requires a signed processing agreement.

Built on open-source software: PDFium (BSD-3-Clause / Apache-2.0, via pypdfium2), Tesseract OCR (Apache-2.0), Pillow (MIT-CMU) and httpx (BSD-3-Clause).

Questions or a feature request? Open an issue on the Issues tab.