Google DeepMind Ships SL2T Sign-Language-to-Text AI in Gboard and Live Transcribe on Pixel 11

Google DeepMind announced on August 12, 2026 that its SL2T (sign-language-to-text) model is now powering sign-to-text dictation in Gboard and Live Transcribe on Pixel 11 phones, marking the first deployment of a massively multilingual sign language AI model directly into consumer products at no additional cost. Starting with American Sign Language to English, Deaf and hard-of-hearing users can sign into their phone anywhere they would normally type — searching the web, drafting messages, or asking Gemini to complete tasks — and sign responses back and forth in Live Transcribe conversations instead of typing. SL2T was trained on more than 100,000 hours of data spanning over 50 sign languages (roughly a quarter of it ASL) and scored 70 BLEURT on the FLEURS-ASL benchmark, well above previously reported systems. For privacy, an on-device model converts camera footage into pose landmark coordinates before anything is sent to Google's servers, and SL2T translates directly from those landmarks into text rather than through intermediate glosses; raw video is discarded immediately after processing. Google said the model is optimized for left-handed and one-handed signers and that it convened an AI Sign Language Advisory Committee (AISLAC) with Deaf organizations and experts to guide ongoing development, with more devices and sign languages planned.