Google Pixel 11 adds ASL sign-to-text for Gboard and Live Transcribe

Google says the Pixel 11 series now supports American Sign Language sign-to-text input, bringing a DeepMind translation model into Gboard and Live Transcribe for people who sign instead of type or speak. The feature was announced on August 21 and is available on Pixel 11 devices, with Google saying users can add sign-to-text to the Gboard toolbar, sign to the front camera, and see English text appear on the outer display.

The practical shift is simple: a phone keyboard can become a camera-based accessibility input. For Deaf and hard of hearing users who communicate in ASL, that may make everyday tasks such as drafting a message, searching the web, responding in Live Transcribe, or asking an AI assistant a question feel closer to natural language use than tapping English on glass.

What changes for users

Google says the feature is powered by SL2T, a multilingual sign-language-to-text model introduced by Google DeepMind. DeepMind said the model was trained on more than 100,000 hours of data across over 50 sign languages, with about a quarter of that data in ASL. The first consumer rollout starts with ASL to English on Pixel 11, while more devices and additional sign languages are planned later.

One privacy boundary matters: DeepMind says an on-device MediaPipe Holistic model tracks pose landmarks, then sends geometric coordinates for translation rather than sending the original camera video stream. That design does not remove every privacy concern because translation still depends on server processing, but it is a meaningful difference from uploading raw video of a signer.

The feature also shows why sign-language AI is harder than a simple word-for-word transcription problem. DeepMind notes that sign languages have their own grammar and use hands, arms, torso, head, and facial movement at the same time. Its team also says it worked on latency, non-signing inputs, left-handed signers, and one-handed signing, which matters when a phone is held in the other hand.

Social hook: if sign-to-text works reliably in public, the phone keyboard stops being only a typing surface and becomes a language access tool.

For CyberOGZ readers tracking mobile AI, this is a useful marker for future coverage: accessibility features are becoming a real test of on-device sensing, cloud translation, privacy design, and product availability. One likely effect is that buyers who depend on assistive technology will judge AI phones less by generic model branding and more by whether these features work accurately in their own language, device, and daily environment.

Sources

• blog.google
• deepmind.google

Cover photo by RDNE Stock project on Pexels, used under the Pexels License.