Retyping text from a photo of a notice, a certificate or a screenshot is slow and error-prone. This tool reads the text for you (OCR) in English, Hindi and Marathi, and gives it back to copy, edit or save as a text file. Recognition runs in your browser, so documents with personal details never leave your device. Lines the reader is unsure of are listed, so you know what to check.
How to use it
- Choose an image: a photo, scan or screenshot (JPG, PNG, WebP, AVIF, BMP or GIF), or up to 20 at once.
- Tick every language in the image. English and Hindi are selected to start with.
- Select "Read text". The first time, each language is downloaded once and then kept on your device.
- Copy the text, or download it as a .txt file. Check any lines listed as unsure.
Options explained
Languages
English, Hindi and Marathi. Tick all that appear: a Hindi notice with English names reads best with both ticked. Hindi and Marathi share the Devanagari script but have their own words, so pick the right one.
Confidence
For each line the reader reports how sure it is. Lines below 60% are listed after reading so you can compare them with the image.
Example
- Input: A phone photo of a printed Hindi certificate, slightly tilted
- Settings: Hindi and English
- Result: The full text, with about 97 of every 100 characters right on our test documents
Limits and troubleshooting
- Printed text reads best. Handwriting is often misread; a higher-accuracy AI reader for handwriting is planned for this tool.
- Small, blurry or very dark text needs a sharper photo, taken straight on in even light. Crop away anything that is not text for a faster, cleaner result.
- Tables and columns come out as plain lines of text, without the layout.
- The first use of each language downloads its data (1 to 3 MB). After that it is read from your device.
Questions
Why choose Marathi instead of Hindi?
Both use Devanagari, but the Marathi model knows Marathi words and letters such as ळ, so it makes fewer mistakes on Marathi text.
Is my image uploaded to read the text?
No. The text reader runs inside your browser. The image and the text stay on your device; our servers never receive them.
How accurate is it?
On our printed test documents, 99 to 100 of every 100 characters were right on clean scans, and about 97 on a tilted, soft phone photo of Hindi text. Accuracy drops with blur, poor light and handwriting, so always check names and numbers.
Can I get text from a scanned PDF?
Turn the PDF pages into images with PDF to JPG, then read them here. Reading scanned PDFs directly is planned.