PDF OCR

When a PDF is really just a picture of text

A scanned contract, a photographed receipt saved as a PDF, an old document digitized years ago: these all look like text on screen, but to a computer they are just images, with no text layer to select, search or copy. Optical character recognition, OCR, reads the shapes of the letters visually and turns them back into real text, the same job a document scanner app does on a phone.

How it works

Each page of the PDF is rendered as an image at a resolution good for reading small text, and a text recognition engine, running entirely on your device, reads the letters and words from that image. The recognized text from every page is combined into one block you can copy, search or download.

The first time takes a little longer

The recognition engine and its language data, a few megabytes, are downloaded the first time you use this tool or the image-to-text tool, and kept by your browser after that. This means the very first page takes longer than the rest while that download finishes; pages after that are quicker.

Getting a good result

OCR accuracy depends heavily on the quality of the scan: a clear, high-resolution, well-lit scan with straight, unrotated text reads far better than a blurry photo taken at an angle. If the source document is tilted, the deskew tool can help before running OCR. Handwriting is read far less reliably than printed text, and some errors should be expected on any source, so always check the result rather than trusting it blindly for anything important.

What it does not do

This gives you the text, not a searchable PDF with the original images and an invisible text layer added back in, which is a more complex file format to build than a browser tool can reasonably produce. It reads what is visible and returns text only. Everything runs in your browser: the images from your PDF are never uploaded anywhere.