Google Pixel 11 Integrates DeepMind Sign-Language-to-Text AI

Google has launched the Pixel 11 smartphone family featuring a new sign-language-to-text (SL2T) model developed by DeepMind. The integration provides deaf and hard-of-hearing users with a more natural modality for digital interaction, bringing sign language processing closer to the utility of spoken language AI.

Integrated directly into Gboard and Live Transcribe, the tool allows users to transcribe sign language into text for web searches, messaging, and interacting with the Gemini AI. At launch, the model specifically supports the transcription of American Sign Language (ASL) into English. To achieve this, DeepMind trained the SL2T model on more than 100,000 hours of multilingual sign language data, with approximately 25% of that dataset consisting of ASL.

A Shift in Translation Logic

While voice-to-text has been a standard smartphone feature for years, sign-language-to-text has remained technically elusive. Most previous systems relied on "glosses"—intermediate written representations of signs—which often failed to capture the nuance of the language. SL2T departs from this method by translating landmarks directly into text. This approach better captures spatial constructions and non-manual markers, such as facial expressions, which are critical to the meaning of sign language.

Privacy and Technical Architecture

To address privacy concerns regarding the use of cameras for transcription, Google implemented a tiered processing system. The device does not send raw video footage to its servers. Instead, an on-device model converts the footage into geometric coordinate wireframes. These anonymized landmarks are then transmitted to the servers for translation, ensuring that the user's visual identity is protected during the process.

Industry Implications

This release targets a global community of over 70 million deaf and hard-of-hearing people who utilize roughly 200 different sign languages. DeepMind stated that deaf people can benefit from sign language processing in the same way hearing people benefit from spoken language processing, noting that the technology opens new possibilities for bridging the communication gap between Deaf and hearing communities.

What's Next

While the current rollout is limited to the Pixel 11 family and ASL, the use of a multilingual training set suggests a roadmap for broader language support. Observers will be watching to see if Google expands the SL2T model to other sign languages or brings the functionality to older Pixel devices and other Android hardware.