Image to Text (Free Online OCR)
xFormat.space extracts text from images with a recognition model running locally in your browser, supporting English, Chinese, Japanese, Korean, French and German without uploading your picture.
Runs in your browser — files never leave your device.
How it works
Turn a screenshot, a photo of a page or a scanned receipt into text you can copy and edit. The recognition model runs on your device, so the image itself is never uploaded. The first run downloads a model of about 15 MB, which is cached for later visits.
- 1
Drop an image (JPG, PNG or WebP) onto the page and open OCR.
- 2
Choose the recognition language that matches the text.
- 3
Click Start Recognition, then edit, copy the text or download it as a TXT file.
Languages and what to pick
There are six choices, selected one at a time: Chinese & English, English, Japanese, Korean, French and German. The Chinese & English option reads Chinese (simplified and traditional) and English together, so use it for mixed text. For a purely English document, the English option is the natural pick. Switching languages loads a different model the first time you use it.
Getting more accurate results
OCR accuracy mostly comes down to the input. Use the original screenshot rather than a compressed copy that was shared through a chat app. For photos of paper, shoot straight on, in even light, with the text filling most of the frame and no shadow across the lines. Crop away busy backgrounds first, and avoid heavy glare. Printed text works best; handwriting and decorative fonts are hit-or-miss.
After recognition you'll see the number of text regions found and an average confidence score. Treat it as a hint, not a guarantee, and proofread numbers, names and anything legal or financial. The text box is editable, so you can fix mistakes before copying. The first-time model download needs an internet connection; your image still stays on your device.
Frequently asked questions
How do I extract text from an image?
Drop the image onto the page, choose the language, and click Start Recognition. The recognized text appears in an editable box that you can copy or download as TXT.
Which languages does the OCR support?
Chinese & English (which also reads traditional Chinese), English, Japanese, Korean, French and German. You choose one language per run.
Is my image uploaded to a server for OCR?
No. The recognition model runs in your browser. Only the model files are downloaded, about 15 MB the first time, and your image stays on your device.
Why is the first OCR run slower?
The first run has to download the text-recognition model and its engine. After that they're cached, so later runs start much faster.
Can it read handwriting?
It is built for printed text. Neat handwriting may partly work, but results are unreliable, so always check the output.