Loading…
Extract text from images, photos, and scans.
Updated
Pull the text out of an image, screenshot, photo, or scanned document using optical character recognition (OCR). Works with JPG, PNG, WebP, GIF, and BMP images. It runs entirely in your browser — your image is never uploaded, so it stays completely private. The recognition engine downloads once on first use, then works offline.
Drag & drop an image, or click to browse
JPG, PNG, or a photo/scan of a document — up to 15 MB.
Best on clear screenshots, documents, and photos of real text. Stylized graphics, logos, and decorative fonts can't be read reliably.
Runs entirely in your browser — nothing is uploaded.
Image to text, or OCR (optical character recognition), is the process of detecting the characters in a picture and turning them into text you can select, copy, search, and edit. This tool runs the open-source Tesseract engine directly in your browser, so the image is processed on your own device and never uploaded to a server. It reads printed English text from JPG, PNG, WebP, GIF, and BMP files up to 15 MB, including phone photos, screenshots, and scanned pages. Accuracy is highest on sharp, well-lit images of standard fonts with strong contrast between the text and its background; handwriting, decorative fonts, and blurry or skewed photos recognize poorly. The engine, a few megabytes in size, downloads once on first use and is cached, so later extractions start instantly and work offline. There is no sign-up and no limit on how many images you can process.
JPG, PNG, WebP, GIF, and BMP, up to 15 MB per image. Any image your browser can display will work. For the best results use a sharp, high-contrast picture of printed text.
No. Recognition runs entirely inside your browser using the Tesseract engine, so the image is never sent to a server and nothing is stored. Once the engine has loaded, the tool even works offline.
Yes. Photograph the page straight-on in good light, keep the lines of text level, and let the text fill most of the frame. Skewed, shadowed, or low-resolution photos lower accuracy noticeably.
Poorly. The engine is trained on printed text, so neat block capitals may partly work, but cursive and everyday handwriting usually do not. For handwritten notes, a dedicated handwriting-recognition service will do better.
Printed English. Words in other Latin-alphabet languages may be partly recognized, but accented characters and non-Latin scripts such as Arabic, Urdu, Hindi, or Chinese are not reliably supported.
The OCR engine, a few megabytes in size, downloads the first time you extract text. Your browser caches it afterwards, so later extractions start immediately.