Camera
Position both hands clearly in frame.
Loading model…
The live predictor improves over time — the model is retrained periodically as the
community contributes signs.
·
Contribute a sign to make it smarter.
Live recognition
Sign in front of the camera — we'll translate it in real time.
The same on-device model the community is training. Predictions refresh several times a second; enable retraining capture if you'd like to contribute your confident clips back to the dataset.
50
Frames per prediction window
1662
Landmarks tracked per frame
0%
Faces stored — landmarks only