Is my image uploaded?

No. The recogniser runs in your browser and the picture never becomes part of a request. The engine is fetched once, from this site, before your file is read.

Clean text on a plain background comes back close to perfect. A photograph of a page at an angle, in poor light, or with text smaller than about twenty pixels tall comes back with mistakes. The confidence figure next to the result is there so you know which you got.

Which languages does it read?

English. That is the language data this page carries, and it is honest to say so rather than to imply the rest and disappoint you after the wait.

Can it read handwriting?

No, and neither can anything else that calls itself OCR. It is built for printed and rendered text, and handwriting comes back as nonsense rather than as an error.

Why is the first run slow?

The recognition engine and its language data are a few megabytes, and they are fetched when you first use the tool rather than on every page view. The second image on the same visit starts straight away.

Does it keep the layout?

Loosely. Line breaks survive, columns and tables usually do not. What comes back is the words in reading order, which is what most people want to paste somewhere.

ON YOUR DEVICE · FREE

Image to Text

Pull the words out of a screenshot or a scan, without the document leaving your machine

…

1
Drop the image in
A screenshot, a photo of a page, a scan. Up to 25 MB. It is read by the page and never sent anywhere.
2
Wait once
The reader is a few megabytes and arrives the first time you use it on a visit. After that it is already there.
3
Copy or download
The text comes back with a confidence figure, because OCR can be wrong in ways that look right.

FAQ

Is my image uploaded?: No. The recogniser runs in your browser and the picture never becomes part of a request. The engine is fetched once, from this site, before your file is read.
How accurate is it?: Clean text on a plain background comes back close to perfect. A photograph of a page at an angle, in poor light, or with text smaller than about twenty pixels tall comes back with mistakes. The confidence figure next to the result is there so you know which you got.
Which languages does it read?: English. That is the language data this page carries, and it is honest to say so rather than to imply the rest and disappoint you after the wait.
Can it read handwriting?: No, and neither can anything else that calls itself OCR. It is built for printed and rendered text, and handwriting comes back as nonsense rather than as an error.
Why is the first run slow?: The recognition engine and its language data are a few megabytes, and they are fetched when you first use the tool rather than on every page view. The second image on the same visit starts straight away.
Does it keep the layout?: Loosely. Line breaks survive, columns and tables usually do not. What comes back is the words in reading order, which is what most people want to paste somewhere.

Image to Text

Drop the image in

Wait once

Copy or download

FAQ

More tools like this one