Extract text from an image
Extract text from an image or a screenshot online and copy it in one click.
or click to pick them from your device
Up to 100 MB per file · You can also paste an image from the clipboard: Ctrl + V
Settings
The language data weighs 3.6 MB: it downloads on the first run and stays in your browser, so the next run starts right away.
No files added yet
Files are processed in your browser and never sent to a server.
How to use it
- 1
Drop an image into the window: a photo of a page, a sign or a screenshot all work.
- 2
Pick the language of the text. If the image mixes two, take a pair — Serbian and English, for instance.
- 3
Press Process. The text appears in the field below the settings: edit it, copy it with one button or save it as a .txt file.
Common questions
Does the image go to a server?
No, recognition runs in your browser. The engine and the language data are downloaded to you, while the image stays on your device — which makes a passport, a contract or a medical record safer to read here than on services you have to upload the file to.
Why is the first run so slow?
Before the first recognition the browser downloads the engine and the data for the chosen language — several megabytes, and they do not arrive instantly. After that everything sits in the browser's storage and the next images are recognised straight away.
The text came out with mistakes — what now?
Usually the text is too small or tilted: the letters have to be clearly distinguishable in the image itself. A larger shot taken straight from above, without glare or a hand's shadow, helps. Check the language too — Latin text read by a Cyrillic model turns into letters that merely look similar.
Does it read handwriting?
No. The model is trained on printed text and turns handwriting into nonsense — that is a different class of problem, not a setting to switch on.
Why are there two kinds of Serbian?
The model recognises letters rather than a language, and Cyrillic and Latin are separate sets of data. Choose by what the image shows: Latin for Beograd, Cyrillic for Београд.
What about text in a PDF?
Split the document into images with the neighbouring tool and bring the page you need here. And if the PDF is built from text rather than scans, recognition is not needed at all — the text is already there and can be selected in any reader.