Image to Text (OCR)
Extract text from images, screenshots and documents using local OCR.
Runs LocallyNo Upload100% Private
How it works
- 1Select an image — it is loaded directly into your browser, never uploaded.
- 2The PaddleOCR detection + recognition models (~7 MB) are loaded once from this site and cached by your browser.
- 3PaddleOCR runs on WebGL on your device: it detects every text box, then recognizes its content.
- 4Copy the recognized text or download it as a TXT / JSON file.
FAQ
Are my images uploaded to a server?
No. OCR runs entirely inside your browser on WebGL. Your images never leave your device.
Which languages are supported?
Simplified Chinese and English in one unified model — mixed Chinese/English text is recognized in a single pass, no language switching needed.
Why load models on the first run?
The OCR models (~7 MB total) are served from this site and cached by your browser. Later runs start instantly, even offline.
How accurate is it?
PaddleOCR is strong on natural scenes, receipts and documents. Blurry or low-resolution photos may need a retry with a sharper source.