Make a scanned PDF searchable
Read the words out of a scan and lay them invisibly over the pages. The document looks the same — it just becomes searchable, selectable and copyable. Pages scanned askew or on their side are set straight.
- Processed on your device
- No upload, no waiting
- No signup, no watermark
How it works
Add your scanned PDF
Drop the file. The tool checks whether it already has a text layer and tells you if OCR would be a step backwards.
Check the language
The tool reads the first page and tells you which language it found. Over 120 are available if you need to pick one yourself.
Download
You get the same document, now with a hidden text layer behind the scan.
Frequently asked questions
Will the pages look different afterwards?
Only straighter: a page scanned a degree off square or on its side is set upright. The scan itself is kept and the recognised words are placed over it in invisible ink — never painted, never printed. What changes is that Ctrl+F starts finding things.
Is my document uploaded?
No. Recognition runs in your browser. The reader and the language pack come from this site the first time you use a language; the document itself never leaves the device.
How accurate is it?
On clean scans of printed text, very. Blurry photos, handwriting and unusual fonts reduce it; skew is corrected before reading. Ruled tables are read cell by cell, so form labels do not run into their values.
Which languages work?
Over 120, including English, Bengali, Hindi, Urdu, Arabic, Spanish, French, Chinese, Japanese and Korean. The script is detected from the page; where several languages share a script, your browser's language is used as the tie-breaker and you can change it.
Some words are missing from the text layer. Why?
The text layer can only carry characters a bundled font can write. Bengali, Devanagari, Arabic, Thai and two dozen other scripts have their own font here; Chinese, Japanese and Korean do not yet, so those words are counted for you rather than saved as blanks. PDF to Word with OCR gives you that text in full.
How long does it take?
A few seconds per page, because your own device is doing the reading rather than a server farm. Progress is shown page by page.
Good to know: Chinese, Japanese and Korean words are left out of the text layer and counted for you, rather than saved as blanks, until a font for them ships; every other script the reader knows is written. Pictures, logos and signatures are ignored so they never become nonsense text. Handwriting is not reliably recognised by any general OCR engine.