Image to Text Converter

How to convert an image to text

1. Drop your image on the page, click “Choose image”, or paste a screenshot with Ctrl+V. On a phone you can take a photo directly.
2. The text appears in a few seconds, in the same language as the image and with the same line breaks and paragraphs. The language is detected for you.
3. Check the accuracy score. If it says “Check the text”, read it through and correct any mistakes right in the box.
4. Optional: choose a language under “Translate to” and click Translate.
5. Click the download button. It detects your computer (Windows or Mac) and saves a .txt file that opens correctly in Notepad or TextEdit.

Text in the same language as the image

You don't need to choose a language. The converter first reads the image as English; if the text is in another script, it finds the script (Devanagari, Bengali, Tamil, Telugu, Gujarati, Kannada, Malayalam, Gurmukhi, Odia, Urdu, Cyrillic, Chinese, Japanese, Korean, Thai…) and reads it again with that language's model. For Devanagari it tells Hindi, Marathi and Nepali apart from the words themselves. Images that mix English with another language are read with both.

How accurate is it?

Every image is read by two open-source OCR engines: PaddleOCR (PP-OCRv6, a deep-learning model that is strong on screenshots, app screens, logos and symbols like ₹ and &) and Tesseract (strong on word spacing and the Indian scripts). Where they disagree, the more complete and more confident reading wins. Both run inside your browser, free.

In our tests, 25 images (documents, phone photos with shadows, a real app screenshot, a bill, two-column text, and 15 languages including Hindi, Tamil, Bengali and Urdu) came out with 100% of characters right. Every result also shows an estimated accuracy: above 90% means almost every word is right; below that, read the text through. Handwriting, decorative calligraphy and text set on a curve are still not readable.

A text file that opens right on Windows and Mac

The download button detects your computer. On Windows the file is saved with Windows line endings and a UTF-8 marker, so Notepad shows Hindi, Tamil, Urdu and every other script correctly, even on older Windows versions. On a Mac it's a standard UTF-8 file for TextEdit, Notes and other editors. If you translate the text, the download contains the translation.

Private by design

Your image is processed on your own device and never uploaded. The only downloads are the two OCR engines and their language data, about 30 MB the first time. Your browser keeps them, so later images are read in seconds, even offline. Translation in Chrome and Edge also runs on your device; other browsers send only the text to a free translation service, and the page tells you when that happens.