Extract text from a scanned PDF with tesseract.js — fully local.
Updated
Up to 200 MB. PDF never leaves your device — pages are OCRed locally.
Quick start
Turn a scanned / image-only PDF into selectable text using pdf.js + tesseract.js, locally.
Drag a scanned PDF onto the drop zone, click to pick it, or paste from the clipboard. The file stays on your device.
Choose the dominant language, render DPI (200 is a good default), and whether to OCR every page or just a range.
Hit Extract text. Each page is rendered locally, OCRed by tesseract.js, and the combined text appears in an editable textarea — copy or download as .txt.
In-depth guide
This tool turns scanned / image-only PDFs into selectable text. Each page is rendered with Mozilla's pdf.js, then the rendered image is run through tesseract.js — a WebAssembly port of Google's Tesseract OCR engine. Everything happens inside your browser; the only network traffic is downloading the language model the first time you use it.
Test first: open the PDF in any reader and try to select and copy a sentence of body text.
DPI controls how detailed each rendered page is before OCR.
Language — pick the dominant script of the document. Mismatched language packs cause cascading misreads (English digits will still work in any pack, but accented or non-Latin text needs the matching language). For CJK content, Simplified and Traditional Chinese are separate packs.
For long documents — or to test settings before committing to a full pass — pick Selected pages and enter ranges:
1-5 → first five pages3,7,11 → those three pages1-3,9,15-18 → mixed rangesThis is how we recommend you test: run a 2-page sample first, check the text quality, then re-run on the full document with confidence.
What in-browser OCR is not good at:
For everything else — printed body text, headings, captions, scanned receipts and invoices in any of the 14 supported languages — accuracy is solid.
No. pdf.js opens the file in your browser, each page is rendered to a local canvas, and tesseract.js OCRs the canvas — all without a network round-trip for your data. Open DevTools → Network while extracting and you'll see zero outbound requests for your PDF.
If your PDF already has a real text layer (you can select + copy text inside a PDF reader), the PDF → DOCX tool is much faster — it extracts the existing text directly. Use OCR only for scanned / image-only PDFs where the page is essentially a picture of text with no underlying selectable text.
Each page has to be rasterised at 150–300 DPI and then run through a WebAssembly OCR engine. Expect 2–10 seconds per page on a modern laptop, longer on mobile. For a 50-page scanned report, plan on a few minutes.
200 DPI is the sweet spot for most pages. 150 is faster but loses accuracy on small text; 300 is best for tiny fonts or noisy scans but doubles the OCR time. If the source PDF was scanned at a higher DPI, rendering above that doesn't help — it just creates bigger images.
Yes. Pick "Selected pages" and enter a range like 1-3,5,7-9. Only those pages get rendered + OCRed.
14 curated languages out of the box: English, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Hindi, Arabic, Japanese, Korean, and both Simplified + Traditional Chinese. The matching trained-data file is fetched from the default Tesseract CDN the first time you use a language and then cached.
Yes — use the dedicated /tools/pdf-searchable tool. It runs the same OCR but overlays the recognised text invisibly back onto the original PDF pages, so the result is a normal-looking PDF where text is selectable, copyable and searchable.
No. OCR output is plain text in reading order — multi-column layouts may interleave, and tables come out as flowing text. For structured extraction you need a heavier pipeline (e.g. AWS Textract or Azure Document Intelligence), which can't run in a pure browser.
No. Decrypt the file first using /tools/pdf-unlock (or any PDF reader's Save-without-password), then OCR it here.
Keep exploring
Hand-picked utilities that pair well with the one you're on — all free, client-side, and zero-signup.
Generate downloadable PNG/SVG QR codes from any string.
Strong random passwords with length and charset controls.
Compress JPEG, PNG and WebP without uploading anywhere.
Extract text from any image with tesseract.js — 14 languages, fully local.