Skip to main content

Extract text from a photo or scan, on your own device

Image to Text Converter reads the words out of a picture and gives them back as text you can edit, copy or search. Point it at a photograph of a page, a screenshot, a scanned document, a receipt or a signboard, and it returns the characters it finds. It recognises around 22 languages, including Hindi and Bengali alongside English.

The recognition runs inside your browser. Your images are not uploaded, which matters when the page you are photographing is an ID, a bank statement or a contract.

What you can do

How to get accurate results

Optical character recognition is only as good as the picture. A sharp, flat, evenly lit photograph taken straight on will read almost perfectly; the same page shot at an angle in a shadow will produce nonsense. Fill the frame with the text, keep the page flat rather than curled, and avoid the shadow of your own phone falling across it. Higher contrast — black on white — reads best.

Handwriting will not work. This reads printed and typed characters; handwritten notes are a different and much harder problem. Expect to correct proper nouns, unusual formatting and tables by hand, and always check numbers yourself before relying on them — a misread digit in an amount or an account number causes real damage, and no OCR system is exempt.

Frequently asked questions

Are my images uploaded?

No. Recognition happens on your device, so the pictures and the extracted text stay with you.

Which languages are supported?

Around 22, including English, Hindi and Bengali. Select the language before scanning — running Devanagari through the English model produces gibberish.

Can it read handwriting?

No, only printed or typed text.

Why is it slow the first time?

The language model has to load before the first scan. After that it is much quicker.

Does it keep the layout?

Broadly, but not exactly. Columns, tables and complex layouts often come out as running text and need tidying.