Putting sign language AI into users’ hands

While AI for spoken languages has advanced rapidly—enabling translation, dictation, and conversational interfaces—the world's 200+ sign languages and an estimated 70 million Deaf and hard of hearing users have been left behind. Sign languages are the primary languages of Deaf communities and a cornerstone of cultural identity, yet progress in sign language AI has been slow due to complex technical challenges and widespread misconceptions about how these languages work.

Today, DeepMind introduces SL2T, a massively multilingual sign-language-to-text translation model that marks a breakthrough in quality and generality. It powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, starting with American Sign Language (ASL) to English, with more devices and languages to follow. Users can sign to search the web, draft messages or documents, ask Gemini to solve queries, and sign responses in Live Transcribe instead of typing. Testers report that signing in ASL is faster, more natural, and more delightful than typing in English.

SL2T overcomes two core challenges. First, sign languages are independent natural languages with their own grammars and lexicons, so the task requires true machine translation rather than a sequential sign-to-word mapping. Second, the model must learn to 'see' and understand physical movement, since sign languages convey meaning through simultaneous movements. This brings sign language AI out of the lab and into consumer products for the first time, opening new possibilities for bridging communication between Deaf and hearing communities.