Google DeepMind unveils sign-language-to-text model to expand access for Deaf and hard of hearing users

Google DeepMind’s Sign Language Team introduced a sign-language-to-text (SL2T) model designed to bring sign language AI out of the lab and into consumer products for the first time. The model powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, starting with American Sign Language (ASL) to English. Additional devices and languages are planned for future releases.

The SL2T model is trained on over 100,000 hours of data spanning more than 50 sign languages. It addresses core challenges in sign language processing, including the need for whole-body movement tracking and full translation rather than sequential sign-to-word mapping. Sign languages are independent natural languages with distinct grammars and lexicons, distinct from spoken languages.

According to internal testing, users reported that signing in ASL is faster, more natural, and more delightful than typing in English. The feature enables Deaf users to sign to their phones in contexts where they would typically type, such as drafting messages, searching the web, or interacting with AI assistants like Gemini.

The announcement highlights the cultural and linguistic significance of sign languages, which are the primary languages of Deaf communities and central to Deaf cultural identity. The technology aims to support access across modalities, acknowledging diverse proficiency levels in signing, speaking, reading, and writing among Deaf individuals.