Extract text from any image without uploading it

How the browser-based OCR works

OCR (optical character recognition) finds shapes that look like letters and turns them into characters. This page uses Tesseract.js, an open-source port of the Tesseract OCR engine that runs in JavaScript. The first time you click Extract text, your browser downloads the engine and the data for the chosen language, about 10 MB in total, from a public code delivery network. After that, it is cached and later runs start much faster.

The recognition itself happens on your computer or phone. Your image is not sent to this site's server or to any AI service, which makes the image to text tool suitable for receipts, letters, notes from meetings and other material you would rather not upload. Processing time depends on your device and the size of the image, usually a few seconds.

Supported languages and when to switch

You can choose English, Spanish, German, French, Portuguese, Italian or Japanese. The menu starts with the language of the page you are on. Choosing the right one matters because each language model knows its alphabet, accents and common words: reading a Spanish menu in English mode will often drop the ñ and the accents, and Japanese text needs the Japanese model to be read at all.

The tool reads one language per run. For a document that mixes languages, run it with the main language first, then again with the second one and combine the parts you need.

Tips for sharp, accurate results

• Use the original: a screenshot of the screen is far better than a photo of the screen.
• Light and contrast: dark text on a light, even background works best. Avoid shadows across the page and glare on glossy paper.
• Hold it straight: photograph the page from directly above so lines are horizontal; tilted or curved lines cause errors.
• Crop to the text: remove busy backgrounds, logos and photos around the text before uploading.
• Make text big enough: small text in a large photo loses detail. Zoom in or crop so the letters are clearly readable to you.
• Printed beats handwritten: Tesseract is designed for printed text. Neat block capitals may work; cursive handwriting usually will not.

If nothing is found, the tool says so; try a sharper image or another language.

What to do with the extracted text

The result appears as plain text with a word count, ready to edit. Always proofread it: OCR commonly confuses similar shapes such as 0 and O, 1 and l, or rn and m, and it does not keep columns, tables or formatting. Common uses include copying a quote from a screenshot, turning a printed handout into editable notes, or extracting text from an image to paste into an AI assistant for a summary or translation; the AI prompt generator can write that summary prompt for you, and the token counter shows how long the extracted text is.

Looking for a description of the picture itself rather than the words in it? The image to prompt tool turns an image into a prompt for image generators.