SL2T brings sign-language dictation to Pixel 11
Google DeepMind has introduced SL2T, a sign-language-to-text translation model designed to support Deaf and hard of hearing users. The system initially powers American Sign Language (ASL) to English dictation in Gboard and Live Transcribe on Pixel 11, with additional devices and languages planned. Users can sign where they would normally type, including when searching, writing messages or documents, and asking Gemini to perform tasks.
SL2T was trained on more than 100,000 hours of data covering over 50 sign languages, with about one quarter of the material in ASL. An on-device MediaPipe Holistic model converts signing into body-pose landmarks; only those coordinates are sent to a server, while the original video can be discarded immediately. The system translates the sequence directly into text rather than using intermediate glosses.
The model achieved a 70 BLEURT zero-shot score on the FLEURS-ASL benchmark, which evaluates ASL-to-English translation. The team also addressed streaming latency, non-signing inputs, left-handed users and one-handed signing. Deaf partners and organizations participated in development, testing and governance through the AI Sign Language Advisory Committee.
Written by Sakhanda Wire AI from reporting by DeepMind. Facts, quotes and full context are in the original article below.