Read the Text From a Scanned PDF

A scanned page is a picture of text, so there is nothing to select or search until something reads the letters back out of the image. PDFdesk does that on your own machine: the recognition engine and its language data are downloaded once to your browser and run there. That is slower than sending the file to a server — expect a few seconds a page, and minutes for a long document — and it is the reason a contract or a medical record never leaves your device. Results come with a confidence score, and pages that scored badly are named so you know where to look.

Why is it slow?

Because it runs on your machine rather than a server. That is the trade for the file never being uploaded.

How accurate is it?

Good on clean printed text, poorer on faint, crooked or handwritten pages. Always check numbers and names.

How can I get better results?

Run Deskew and Improve a Scan first. Straight, high-contrast pages recognise far better.

Which languages are supported?

English, French, German and Spanish. Each needs its own data file, downloaded the first time you use it, so the list only contains languages that really work.