Pull the text out of an image
Read words out of a screenshot, a scan or a photo and get editable text back, in over 100 languages. It runs in your browser, so the image and the text never leave your device. Honest up front: great on printed text, weak on handwriting.
Drop image here or click to upload
JPG, PNG, WEBP, paste from clipboard works too
Cmd+V to paste a screenshot
What it reads well, and what it doesn't
Printed and typed text
Screenshots, typed documents, slide text and clean scans come through reliably. This is the job it is built for.
Photos of text
A photo of a sign or a page works if it is sharp, straight and well lit. Glare, skew and low resolution all hurt.
Handwriting
Cursive and handwriting are read poorly. Expect a rough draft at best, not a clean transcription.
Small things that lift accuracy
Feed it sharp, high-contrast text that runs straight across the image, and tell it the right language. If a photo is tilted, straighten it with the Rotate tool first; if there is a lot of clutter around the text, trim it with the Crop tool. Dark text on a light background reads best. Do those few things and clean printed text usually comes back almost perfectly.
Frequently asked questions
Very good on clean printed text: screenshots, typed documents and decent scans usually come through at well over 90 percent. It drops off on low-resolution captures, skewed photos and especially handwriting, which it reads poorly. Clear, straight, high-contrast text gives the best result.
Not really. The engine is built for printed and typeset text, so cursive and handwriting are often below half right. Treat any handwriting result as a rough draft to fix by hand.
The OCR engine and the language data download once on the first extraction. After that it is cached, so the next ones are quick.
Over 100, including English, Spanish, French, German, Chinese, Japanese, Arabic, Russian, Hindi and many more. Picking the right one makes a real difference to accuracy.
Not directly. Convert the PDF to images first with the PDF to JPG tool, then run each page through here.
No. The OCR runs in your browser with Tesseract.js. The image and the text it reads never reach a server.