Google DeepMind announced the launch of SL2T, a multilingual sign-language-to-text AI model designed for the estimated 70 million people worldwide who are deaf or hard of hearing. The model debuts on the Pixel 11 smartphone, powering a new “sign-to-text” dictation feature on Gboard and Live Transcribe. Initially capable of translating American Sign Language (ASL) to English text, it is the first of its kind in a real-world consumer product. SL2T allows users to sign into their smartphone’s camera to perform tasks such as web searches, drafting emails, and prompting Gemini. The model was trained on more than 100,000 hours of sign language data spanning over 50 languages, with about a quarter of that dataset made up of ASL communications. Google trained it across multiple sign languages to learn shared structural patterns, enabling it to outperform earlier models. For privacy, SL2T uses an on-device computer vision model called MediaPipe Holistic, which tracks geometric pose locations on the signer’s face, hands, arms, and torso, sending only coordinates to the cloud rather than video. The model translates sequences directly into text, bypassing intermediate text annotations known as “glosses,” to better capture non-manual expressions and spatial grammar structures characteristic of ASL.