DeepMind’s new SL2T model brings multilingual sign‑language‑to‑text capability to Pixel 11, enabling real‑world sign‑language dictation on Gboard and Live Transcribe.
Google has unveiled SL2T, a groundbreaking AI model from DeepMind that can translate sign language into text across multiple languages. The technology is being integrated into the Pixel 11’s Gboard and Live Transcribe, promising real‑time sign‑language dictation for users worldwide.
What is SL2T?
SL2T stands for Sign Language to Text. It leverages a large multimodal transformer trained on a diverse dataset of sign‑language videos and corresponding transcriptions, enabling it to recognize and convert gestures from dozens of sign languages into written text.
Multilingual capabilities
Unlike earlier prototypes that focused on a single language, SL2T supports a range of sign languages, including American Sign Language (ASL), British Sign Language (BSL) and several others. This multilingual approach aims to bridge communication gaps for deaf and hard‑of‑hearing communities globally.
Integration with Pixel 11
The model is baked into the Pixel 11’s software stack, allowing users to activate sign‑language dictation directly from Gboard or Live Transcribe. When a user signs in front of the phone’s camera, SL2T processes the gestures and displays the corresponding text in real time.
- Instant transcription of sign language during video calls
- Improved accessibility for messaging apps
- Support for multiple sign languages in a single device
Potential impact
By bringing sign‑language recognition to a mainstream smartphone, Google hopes to make digital communication more inclusive. The technology could also serve as a foundation for future applications such as sign‑language‑enabled virtual assistants and educational tools.