Copy the text from an image

How it works

1. Drop a screenshot, a scan or a photo of a printed document. Up to three images can be read in one run.
2. Press Extract text. The first time, the text recognition models and their runtime (about 11 MB) are downloaded.
3. Check the text in the box and press Copy, or download it as a .txt file.
4. For the best results, photograph the page straight on in even light, so the lines or columns run straight.

Frequently asked questions

Which text can it read?

Printed Japanese, English, German, French, Spanish, Italian, Portuguese and Turkish: documents, receipts, invoices, screenshots of apps and web pages, and vertical Japanese (tategaki) such as novels and letters, which is recognised automatically. Handwriting may not be read correctly.

How accurate is it?

On our test pages of printed Japanese documents, fewer than 1 character in 100 was wrong on scans, screenshots and phone photos taken straight on, and about 2 in 100 on vertical pages, mostly 、 and 。 left out. Pages in German, French, Spanish, Italian and Portuguese also stayed under 1 in 100; Turkish was about 1 to 2, mostly ı and İ read as i or I. Symbols such as ※ and ■ are often dropped too, and horizontal pages set in several columns can come out with the columns mixed, so check the result before you use it.

Is my image uploaded?

No. The recognition models run in your browser, so the picture and its text stay on your device. The models are downloaded once; after that, reading starts straight away and works offline.

How long does it take?

A page full of text takes about 5 to 10 seconds on a fast computer and longer on older laptops and phones. A screenshot with a few lines takes a second or two.

What are the free limits?

The free plan handles 3 images per run, 20 MB each and 10 runs per day. Pro removes these limits.