Recognise the text in a scan, add an invisible searchable layer, extract the words as plain text, and then edit them in place — all inside your browser.
A scanned page is a photograph. There is no text in the file — which is why searching finds nothing, copying gives nothing, and no editor can change a word. OCR looks at the pixels, works out which characters they represent, and records where each one sits.
The result is written back as an invisible text layer: the same recognised words, drawn transparently at the positions they occupy in the image. The page looks untouched, but every tool that reads PDFs — search, indexing, copy and paste, screen readers — suddenly works.
Resolution matters most: OCR is run at 300 dpi here because that is where recognition accuracy plateaus for body text. Beyond that, the biggest wins are practical — scan straight rather than at an angle, avoid shadows across the page, and pick the correct language. A Turkish document read with the English model will mangle every ş and ğ.
Handwriting is out of scope; OCR reads print. Heavy noise, very low contrast, decorative fonts and tightly packed tables all degrade accuracy. And OCR never guarantees correctness — always read the recognised text before relying on it for anything that matters.
Text you can select is text you can search — and text you can edit. The editor exposes the recognised lines to the Edit text tool, and find-and-replace works across the whole document. To black something out permanently, use redaction; to shrink a bulky scan, see PDF compression.
No. Recognition runs inside your browser through an OCR engine compiled to WebAssembly. The only thing downloaded is the language model — a few megabytes, served from this site rather than a third-party CDN, fetched once and cached. Your pages never travel anywhere.
The recognised words are written back into the file as an invisible text layer positioned over the scan. The page still looks exactly the same, but you can select, copy, search and index the text — which is what makes a scan usable.
English, Turkish, German, French, Spanish, Italian, Dutch, Portuguese, Russian and Arabic, plus combined English+Turkish and English+German models for mixed documents. Pick the right one — accuracy drops sharply with the wrong language.
On a clean 300 dpi scan of printed text, very accurate. On a phone photo at an angle, in poor light, or on handwriting, much less so — OCR reads print, not penmanship. Straighten and crop the page first if you can.
Yes. Once a page has been recognised, the Edit text tool treats the OCR lines exactly like a real text layer: click a line and rewrite it. The replacement is drawn over the scan.
Recognition is real work — a page at 300 dpi takes a few seconds on a typical laptop, single-threaded inside a browser sandbox. Run one page first to check the language and quality before committing to a long document.