Fully offline
The model and WASM run in the page. Recognition does not need the network after load.
Free on-device OCR that never leaves this computer
Detection, orientation, and recognition run locally. Nothing is uploaded.
Recognition stays in your browser. Images are not sent to a server.
Offline OCR workbench
Why use Free OCR
Built for text in images
A DET → CLS → REC PaddleOCR pipeline for printed text and screenshots.
Privacy first
IDs, chats, and scans can stay on this device instead of a cloud OCR API.
Load once
After the engine starts, you can recognize more images without re-uploading.
Three steps
- 01
Choose or drop an image
- 02
Read the text in the browser
- 03
Copy the result — the image stays here
FAQ
Are images uploaded?
No. Recognition runs in a WebAssembly engine in your browser.
Why is the first run slow?
The OCR model and WASM need to initialize. Later runs are much faster.
Does handwriting work?
This demo is strongest on print and screenshots. Handwriting is weaker.
Can I use it offline?
The first visit downloads the page. After that, cached files can run offline.