Google DeepMind Launches SL2T: Sign Language AI Comes to Pixel 11
Google DeepMind has introduced a new model for translating sign language into text, known as the multilingual sign-language-to-text (SL2T) system. This innovation integrates sign language AI into consumer applications such as Gboard and Live Transcribe on the Pixel 11, initially focusing on American Sign Language (ASL) to English. Deaf individuals can now communicate via sign language with their phones for tasks like web searches and messaging. The SL2T model was developed using over 100,000 hours of data from 50 different sign languages, interpreting sign language through pose landmarks while ensuring user privacy by omitting video. It achieves an impressive zero-shot BLEURT score of 70, exceeding previous records. The development process included insights from the Deaf community, facilitated by the AI Sign Language Advisory Committee (AISLAC). SL2T is offered on the Pixel 11 without additional charges, with future plans for more languages and devices.
Key facts
- Google DeepMind introduces SL2T, a sign-language-to-text translation model.
- SL2T powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, starting with ASL to English.
- The model is trained on over 100,000 hours of data across more than 50 sign languages.
- SL2T achieves a zero-shot score of 70 BLEURT on FLEURS-ASL benchmark.
- The model uses pose landmarks instead of raw camera feed for privacy.
- SL2T translates directly from landmarks, bypassing glosses.
- The AI Sign Language Advisory Committee (AISLAC) was established for responsible deployment.
- The feature is available at no additional cost on Pixel 11, with more devices coming soon.
Entities
Artists
- Sam Sepah
Institutions
- Google DeepMind
- Android
- Gboard
- Live Transcribe
- Pixel 11
- Gemini
- MediaPipe Holistic
- AI Sign Language Advisory Committee (AISLAC)