How to use OCR PDF
- 1
Add the scanned PDF
Drop a file from a scanner or a phone scanning app; the tool checks how many pages already carry selectable text.
- 2
Pick the languages
English is preselected and your browser's language is suggested; add up to four languages that really appear in the document.
- 3
Set accuracy and pages
Best reads pages at 300 DPI and Faster at 200 DPI. Limit the run to a page range, and skip pages that already have text.
- 4
Download the results
Save the searchable PDF, or download everything recognized as a .txt file with a marker before each page.
Limits and Pro
Free: PDFs up to 50 MB, OCR on up to 200 pages per run, and up to 4 recognition languages at once out of 45 available.
Pro raises the cap to 5,000 pages per run and files up to 4 GB; recognition speed still depends on your device's processor. See Pro
About OCR PDF
The result is your original PDF with an invisible text layer added to each scanned page. The scan itself is not re-compressed or redrawn, so it looks exactly as before, but every recognized word now sits at its position on the page: Ctrl+F finds it, click and drag selects it, and copy and paste produces real text. Pages are rendered at 300 DPI (200 in Faster mode) and read by Tesseract, the open-source OCR engine, running as WebAssembly in your browser, with several pages processed in parallel and a status line for each one.
Pages scanned sideways or upside down are detected and read the right way round, pages that already have selectable text can be skipped so mixed documents aren't processed twice, and the text layer is stored as Unicode, so Cyrillic, Greek, Arabic, Hebrew, Hindi, Thai, Chinese, Japanese and Korean results can be searched and copied just like Latin script. A plain-text export collects everything recognized, page by page.
Typical uses: making archived letters and contracts searchable, pulling figures out of a scanned invoice, preparing a scan for conversion to Word, or giving screen-reader users text they can hear.
Accuracy depends on the scan. Clean printed text at 300 DPI is usually read almost perfectly; handwriting, low-contrast photos, decorative fonts, tiny print and pages photographed at a steep angle produce more mistakes. Strongly tilted pages are not straightened, and the engine reads words rather than layout, so columns and tables come out as plain lines of text in the export.
Privacy
Scanned pages are rendered and recognized inside this browser tab by Tesseract compiled to WebAssembly; neither the PDF nor the recognized text is sent to a server. The engine and the language files you pick, 2 to 15 MB each, are downloaded once from a public CDN and cached by your browser.
Features
Search and copy
Recognized words are positioned over the scan, so find, select and copy work in Acrobat, Preview, Chrome and other readers.
Scan quality untouched
Only an invisible layer is added, so stamps, signatures and photos keep their original sharpness and file size barely grows.
45 languages
From English, Spanish and German to Arabic, Hindi, Chinese, Japanese and Korean, combinable up to four at a time.
Rotated pages
A sideways or upside-down page is retried in each orientation and the best reading is kept.
Mixed documents
Pages with an existing text layer are left alone, which saves time on PDFs that are only partly scanned.