PDF to Text - Extract Text from Any PDF, with OCR for Scans

One tool for digital and scanned PDFs

"Getting the text out of a PDF" is really two different problems. A digitally created PDF (exported from Word, generated by an invoice system) carries a machine-readable text layer that can be read out directly and losslessly. A scanned PDF is just pictures of pages - the text has to be recognised with OCR, which is slower and imperfect. Most sites make you figure out which kind you have and pick the right tool. This one checks for you: on upload, every page is probed for a text layer with PDF.js. Pages with real text are extracted instantly; pages without are flagged and queued for a local Tesseract OCR pass that you trigger with one click.

Mixed documents are handled per page

Real-world documents are often hybrids: a digital contract with a scanned signature page appended, or a report where one chapter was photocopied back in. Because detection runs page by page, the fast path and the OCR path combine cleanly - digital pages come from the text layer, scanned pages from OCR, and the final output stitches them together in page order. Each page in the result is labelled with its source, and OCR pages show a confidence score so you know which ones to proofread.

Privacy: nothing leaves your browser

Both extraction and OCR run locally. PDF.js parses the file in your browser's memory, and Tesseract runs as a WebAssembly module on rendered page images. The only network request the tool ever makes is downloading the ~4 MB Tesseract language data file from a CDN the first time you run OCR. Your document itself is never transmitted anywhere - which matters when the PDF is a contract, a medical record, or bank correspondence.

Output formats

Plain text (.txt) and Markdown (.md) are always available. After an OCR run three more formats unlock: a searchable PDF (your original scan with an invisible text layer embedded at each recognised word, so it becomes selectable and searchable in any viewer), hOCR (Tesseract's HTML format with per-word bounding boxes), and JSON bounding boxes for programmatic processing. The Markdown conversion is heuristic - it infers headings from font-size hints - so review it on complex layouts.

Tips for better OCR results

Select the correct document language before running OCR; it is the single biggest accuracy factor. Rotate sideways scans with the hover controls first - Tesseract expects upright text. And if a PDF already has a bad text layer (garbled output from someone else's OCR), click those pages to force a fresh OCR pass; the new result replaces the garbage text.