OCR a scanned PDF
Recognize the text in a scanned PDF and download a searchable copy: the pages look identical, but you can now select, copy and search the text. You also get a plain-text export. Recognition runs entirely in your browser — most OCR sites upload your documents; this one cannot.
100% private: files are processed in your browser and never uploaded to any server.
How it works
- Drop a scanned PDF into the box above and pick the document language.
- Click “Recognize text” — the engine loads once, then processes each page on your device.
- Download the searchable PDF, the plain text, or both.
The difference between a scan and a document
A scanned PDF looks like a document, but to a computer it is just a photograph of one. There is no text inside — only an image of text — which is why you cannot select, copy or search it. OCR (optical character recognition) reads those images the way you would, recognises the letters, and adds a real text layer underneath. The page looks identical, but now it behaves like a proper document.
What you get back
Two outputs. The first is a searchable PDF: your original pages with an invisible text layer added on top, so you can select and search the text while the page still looks exactly as scanned. The second is a plain-text export of everything the OCR recognised, ready to paste elsewhere. The searchable PDF is what makes scanned archives usable — you can finally find a document by searching its contents.
OCR that runs on your device
This is unusual and it matters. Almost every OCR service uploads your documents to a server to process them, which is a real problem given that the documents people most often need to OCR — contracts, medical records, historical papers — are exactly the private ones. Pdfect runs a WebAssembly OCR engine directly in your browser, with the language models served from this site and cached for reuse. Your scans are never uploaded, and after the first run OCR even works offline.
Getting the best accuracy
OCR quality depends heavily on the input. Clean, straight scans at 150 DPI or more read very accurately; skewed, low-resolution, faint or handwritten pages are harder, since OCR reliably reads printed text, not handwriting. Choosing the correct language — or both languages for a bilingual document — also improves results noticeably. If a scan is faint, rescanning at higher contrast before OCR pays off.
Frequently asked questions
What does OCR actually do to my PDF?
It adds an invisible text layer on top of each scanned page. The page image stays exactly as it was, but PDF readers can now select, copy and search the text underneath the image.
Which languages are supported?
English and Spanish, individually or combined for bilingual documents. The recognition models are served from this site and cached — after the first use, OCR even works offline.
Why is it slower than other OCR sites?
Because nothing is uploaded: your own device does the work. A typical page takes a few seconds. In exchange, confidential documents — contracts, medical records — never leave your machine.
How accurate is the recognition?
On clean scans at 150 DPI or more, accuracy is comparable to commercial tools. Skewed, low-resolution or handwritten pages reduce accuracy — OCR engines only reliably read printed text.