Local processing · No uploads
PDF Text & OCR
Hybrid: extract text layers first; OCR pages with no text. Choose OCR-only if an existing text layer is incomplete or incorrect.
Some pages have no text layer. Choose Hybrid or OCR mode to recognize scanned pages.
Limits: 100 MiB per PDF, 500 pages, 5 million output characters. Up to 30 OCR pages per task; OCR-only mode accepts up to 30 pages. Rendering is sequential, capped at 8 million pixels per page.
PDF.js, fonts and, when OCR is needed, the recognition engine and language models download from this site. Models may be cached in this browser. Your PDF and results are never uploaded or saved automatically.
How to use
Choose a PDF and a mode, then process. Hybrid extracts existing text layers and uses OCR for pages without text. Text-only skips OCR; OCR-only recognizes every page. Copy or download UTF-8 TXT. Cancellation and errors retain completed pages, with unfinished downloads marked partial.
Will the reading order be exact?
Text extraction preserves PDF text order and infers spacing. Hybrid does not merge image text into a page that already has a text layer; use OCR-only for such pages or incorrect layers. OCR can misread characters. Columns, tables and handwriting may need corrections. Layout and images are not preserved.
Password-protected documents
This first release does not open PDFs that require a password. Unlock a copy locally using software you trust before extracting. This tool does not execute embedded PDF JavaScript or open document links.