Google DeepMind launched SL2T, a new AI model for real-time sign language translation. The feature debuts on the Pixel 11 series through Gboard and the Live Transcribe app. Users sign into the device camera to perform web searches, send messages, or interact with the Gemini assistant.
The model initially supports American Sign Language (ASL) translation into English. Google trained the multilingual system using more than 100,000 hours of sign language data.
An on-device model converts gestures into geometric coordinates to protect user privacy. The system discards raw video and only transmits coordinates for translation. Google plans to expand support to additional sign languages in the future.