PDF Text Extractor (No OCR)

Extract the text from PDF files at links you supply and get one row per PDF. It's for analysts, researchers, legal and compliance teams and developers who need the text of reports, filings, manuals or papers for search, summaries or AI pipelines.

This is the fast text-only option. It reads the PDF's text layer and does no OCR. For scanned PDFs (pages that are images), use our pdf-ocr Actor.

Page limit: only the first 200 pages of each PDF are read. Longer files still return one row with the text of pages 1 to 200, and the row says so (truncatedAtPages: 200, pagesRead: 200, pages = the file's full page count). Pages after 200 are not extracted.

Input: direct links to PDF files.

Output: the text of up to the first 200 pages (or text per page), page count, pages read, title, PDF version, word and character counts and file size.

Use it only on PDFs you have the right to process.

PDF text input

• pdfUrls: direct PDF links, up to 1,000. Each file is downloaded once.
• splitPages: return pageTexts (one string per page) instead of text
• keepLayout: keep columns and table spacing (off by default: reading order, smaller output)
• maxResults: stop after this many extracted PDFs

Links on example.com are the example input: each returns one free EXAMPLE_INPUT row, nothing is fetched and nothing is charged. Replace them with your own links.

Output fields

url, finalUrl, fileBytes, pages (full page count), pagesRead (pages actually read, at most 200), truncatedAtPages (200 when the file has more than 200 pages and the rest was not read, otherwise null), title, pdfVersion, charCount, wordCount, truncated, extractedTextBytes, text or pageTexts

Text longer than 1,000,000 characters is cut off and flagged truncated. The 200-page cap is separate and is flagged with truncatedAtPages.

Free rows (not charged):

• protected: the PDF is encrypted or password-protected. It is skipped, never decrypted.
• no-text-layer: no text found (probably a scanned PDF). Try our pdf-ocr Actor.
• error: the link failed. Reasons include robots.txt disallowing it, HTTP 401, 403 or 404, a login, cookie or paywall page instead of a PDF, a file over 50 MB, or an unreadable PDF. Failed links are not retried.
• invalid-input, free-plan-limit, charge-limit-reached.

Example: {"pdfUrls":["https://www.irs.gov/pub/irs-pdf/p17.pdf"]}

Data source and compliance

• Fetches only the links you supply. It doesn't crawl or follow links (apart from HTTP redirects, at most 5).
• Checks each site's robots.txt first. If robots.txt disallows the file, or can't be read, the file is not fetched.
• Sends a user agent that names this Actor. No cookies, no logins, no passwords. A 401, 403 or paywall is reported as a failure and never retried.
• Encrypted or password-protected PDFs are skipped and marked protected. There is no decryption and no DRM bypass.
• Reads at most the first 200 pages of each file (hard cap, flagged truncatedAtPages: 200).
• Files up to 50 MB. Each file is held only while its text is extracted, then deleted. Nothing is stored outside the run's dataset.
• Runs default to 1 GB of memory, which also buys more CPU for long documents. Text extraction stops after 5 minutes per file; a file that hits the limit is an error row and is not charged.
• Text is extracted with poppler's pdftotext (GPL), run as a separate program.
• Free-plan accounts get up to 100 results per run.

Pricing. Pay per result: $0.80 per 1,000 PDFs ($0.0008 each), the same price on every plan, no start fee. Free status rows are never charged.