About
SL2T (Sign-Language-to-Text) is a machine-translation model family developed by the Google DeepMind Sign Language Team. The models translate sign language into spoken-language text from a sequence of pose landmarks describing the hand, face and body movements of a signer. The first model, SL2T 1.0, was unveiled on 12 August 2026; it translates American Sign Language (ASL) into English and was deployed in the Gboard and Live Transcribe apps. Development is guided by the AISLAC advisory committee.
