Image to text

Text you can copy, from any picture

A receipt you photographed, a slide in a screenshot, a page of a book, a sign: the words are in the picture, but you can't select them. Optical character recognition (OCR) reads the letters and gives you the text. Choose the picture, pick the language and press Extract text. The text appears on the page with a Copy button, and you can download it as a .txt file.

Tips for good results:

• Use a sharp picture taken straight on, with even light and no shadows across the text.
• Make sure the text is large enough: about 20 pixels tall or more works best.
• Pick the document's language; accents and special letters (ç, ã, ñ) are read correctly only with the right one.

OCR reads printed text well. Handwriting and decorative fonts are much less reliable.

Private by design

Recognition runs in your browser with Tesseract, the open-source OCR engine. Your picture isn't uploaded; the language file is downloaded once (about 1 to 3 MB) and then kept by your browser.

The picture is decoded in a Web Worker and passed to Tesseract, compiled to WebAssembly, using its neural-network (LSTM) recogniser and the language model you chose. Tesseract finds text lines, segments them into words and recognises the characters, then returns the text in reading order. Everything, including the engine and language files, is served by UtilNest and runs on your device.