Extract Text from Image
Pull the text out of any image — photo, scan or graphic — with an OCR engine that runs on your own device. Twenty languages, editable output, one-click copy and TXT download. Nothing is uploaded.
Privacy-first: every image is processed locally in your browser. Nothing is uploaded, stored or transmitted to any server.
From pixels to text you can use
Text trapped in an image can’t be searched, quoted or pasted — retyping it is slow and error-prone. OCR reads the pixels for you: drop the image in, and seconds later the words sit in an editable box with a confidence score, ready to copy or save as a .txt file. Selecting the right language (up to three at once for mixed documents) is what makes accuracy high — all 20 site languages are available.
The engine runs where your documents are
This tool uses Tesseract — the most widely used open-source OCR engine — compiled to WebAssembly so it runs inside the browser tab. The first run fetches the engine (≈5 MB) and your language pack (1–3 MB), tells you so beforehand, and caches both. From then on, extraction works instantly and even offline: your contracts, letters and IDs are read on your machine, never on a server.
Frequently asked questions
- Is my image uploaded to extract the text?
- No. Recognition runs on your own device using a WebAssembly build of the Tesseract OCR engine. Documents, IDs, letters and screenshots never touch a server.
- Why is there a one-time download?
- The OCR engine (≈5 MB) and a trained data pack for each language you select (≈1–3 MB each) have to reach your browser once. You are told the size before anything downloads, and everything is cached locally — later runs work instantly, even offline.
- Which languages are supported?
- Twenty: English, Spanish, Portuguese, French, German, Italian, Dutch, Polish, Ukrainian, Russian, Turkish, Arabic, Hindi, Indonesian, Vietnamese, Japanese, Korean, Chinese (Simplified and Traditional) and Estonian. You can select up to three at once for mixed-language images.
- How accurate is it?
- On clean screenshots and scans, very accurate. Photos work best when the text is sharp, well-lit and roughly horizontal. A confidence score is shown with every result, and the output is editable so you can fix stray characters before copying.
- What can I do with the result?
- The extracted text appears in an editable box — copy it with one click or download it as a .txt file named after your image.
- Can it read handwriting?
- No — like most OCR engines it is built for printed and on-screen text. Neat block capitals sometimes work, but cursive handwriting is not supported.