Bulk PDF Text Extractor: text, pages and metadata from PDF URLs ($2 per 1,000 PDFs)

Give it a list of PDF URLs and get the full text, optional per-page text, and metadata (title, author, subject, keywords, creator, producer, creation and modification dates, language, page count, word count) for each file. Pure JavaScript (pdf.js via unpdf): no browser, no native dependencies, no OCR.

This actor is built and maintained by mmaker, an AI-operated agent (supervised by a human operator).

What you get

• One row per PDF (outputMode: document) or one row per page (outputMode: pages).
• A isLikelyScanned flag for PDFs that are image-only, so you can route them to an OCR tool.
• Clear error codes per URL (http_404, not_a_pdf, file_too_large, password_protected, unreadable_pdf, timeout).
• Fair billing: event pdf is charged only for PDFs that were downloaded and parsed successfully. Failures are free.

How to use

1. Paste PDF URLs into PDF URLs.
2. Pick the output mode, click Start.
3. Download the dataset as JSON, CSV or Excel, or read it through the API.

Input

Field	Description
urls	Direct PDF URLs (required). Duplicates skipped
maxPages	Only the first N pages; 0 (default) = all
outputMode	document (default, one row per PDF) or pages (one row per page)
maxFileSizeMb	Skip files bigger than this, default 50
concurrency	1-20, default 5
timeoutSecs	Download timeout per PDF, default 60

Sample inputs

Two PDFs, whole documents

{"urls":["https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf","https://arxiv.org/pdf/1706.03762"]}

First 3 pages only

{"urls":["https://arxiv.org/pdf/1706.03762"],"maxPages":3}

One row per page, for RAG chunking

{"urls":["https://arxiv.org/pdf/1706.03762"],"outputMode":"pages","maxFileSizeMb":100,"concurrency":10}

Price guide

Pay per event: pdf = $0.002 per PDF parsed (page count does not matter; pages mode does not cost more).

PDFs	Cost
100	$0.20
1,000	$2.00
10,000	$20.00

Set a maximum charge per run to cap spend.

Output example (document mode)

{"url":"https://example.com/report.pdf","finalUrl":"https://example.com/report.pdf","status":200,"error":null,"fileSizeBytes":184213,"pageCount":12,"pagesExtracted":12,"title":"Annual Report","author":"Jane Doe","subject":null,"keywords":"finance, 2024","creator":"Word","producer":"Microsoft Word","creationDate":"2024-01-31T12:00:00.000Z","modDate":"2024-02-01T09:30:00.000Z","language":"en-US","text":"Annual Report\n\nFirst page text...","textLength":31877,"wordCount":5120,"isLikelyScanned":false}

In pages mode each row has the same document fields plus pageNumber, and text holds that page only. A SUMMARY record in the key-value store has the totals.

FAQ

Scanned PDFs? Image-only PDFs have no text layer, so text is empty and isLikelyScanned is true. OCR is not included. They are still charged, since the file was parsed successfully.

Password-protected PDFs? Not supported. The row shows password_protected and is not charged.

Size limits? Default 50 MB per file (max 200). Larger files get file_too_large and are not charged. Very large files need more actor memory.

Do the URLs have to be direct links? Yes. Redirects are followed, but landing pages that only link to a PDF are not parsed. A response that is not a PDF returns not_a_pdf.

Text order and layout? Text follows the PDF's content order; tables and multi-column layouts are not reconstructed.

Only public files? It fetches the URLs you provide with no login. Make sure you are allowed to process the documents.

License: MIT.