Extract text from images.

No. Text recognition runs entirely in your browser on your device. Your images and the extracted text are not sent to our servers or to any other service.

English and Albanian (including ë and ç), plus a mixed Albanian + English option, have been tested on photos, screenshots and scans. German, French, Spanish, Italian and Portuguese are also available and work well on clean printed text. Choose the language of your text for the best result.

The first time you use a language, its text model (about 1 to 3 MB) and the recognition engine download from this website. They are then kept in your browser, so later extractions start faster, even for new images.

On clear printed text and screenshots it is usually very accurate, but no text recognition is perfect. Blurry photos, very small text, unusual fonts and strong shadows can cause mistakes. Each result shows a confidence score, and you can edit the text before copying or downloading it. Always check names, numbers and amounts against the image.

Handwriting is not reliably supported; results are often incomplete. Pages with two or more columns, such as newspapers, may be read across the columns in the wrong order. For those pages, crop each column into its own image first.

Slightly tilted pages are straightened automatically, and a sideways or upside-down image is usually detected and read the right way round. If a result still looks wrong, use Rotate and read again.

JPG, PNG, WebP and HEIC (iPhone) photos, up to 20 images at a time and 30 MB per image. Animated images and PDFs are not supported.

Yes. Add up to 20 images and they are read one after another. You can stop at any time, copy or download each result, or copy and download all the text together as a single TXT file.