Sign language becomes a native phone input

Pixel 11 will let people sign in American Sign Language and receive English text directly in Gboard and Live Transcribe. The feature is powered by Google DeepMind’s SL2T, or sign language-to-text, model. Its significance is not simply another camera demo: the output enters Android’s everyday keyboard and transcription tools, reducing the need for Deaf and hard of hearing users to switch first to speech, typing, or a separate translation application.

“Starting with American Sign Language-to-English on Pixel 11, people can sign directly into Gboard and Live Transcribe instead of typing.”

Google DeepMind described the initial ASL-to-English deployment in the source post. In follow-up posts, the lab said SL2T reaches state-of-the-art results on academic benchmarks and is designed for practical conditions such as signing with one hand while the other holds a phone. That detail matters because systems trained around two hands, fixed framing, and controlled backgrounds can perform well in a benchmark yet fail as mobile interfaces.

A split between on-device tracking and cloud translation

The processing architecture draws an important privacy boundary. Pixel performs body-pose tracking on the device, while Google’s servers translate those pose representations into text. This avoids describing the system as fully on-device, and it raises concrete questions for the production release: exactly what representation leaves the phone, how long it is retained, what happens on a poor connection, and whether users can review or delete associated data.

Google DeepMind’s official technical overview says the model was developed with guidance from Deaf Googlers and an AI Sign Language Advisory Committee. The account normally covers Gemini research, scientific models, robotics, and work transferred into Google products. Here, the notable shift is from a research result to an operating-system input layer. Gboard integration could make the same capability available across messaging, search, and document applications rather than confining it to a single accessibility tool.

What to watch after launch

The first release covers one direction—ASL into English—on Pixel 11. The next tests are broader language coverage, regional variation within sign languages, continuous signing across longer context, and end-to-end latency. Real-world error rates will matter more than a general state-of-the-art label, particularly under uneven lighting, partial framing, different signing speeds, and varied mobility. Google says it intends to expand to more sign languages and applications but has not supplied a timetable. Availability, offline behavior, and the quality of fallback controls will show whether SL2T becomes dependable infrastructure or remains a narrow first-generation feature.

#pixel-11

Google SL2T Turns ASL Into English Text on Pixel 11