Google DeepMind Launches Sign Language Translation on Pixel 11

Google DeepMind has released its first consumer-facing sign language AI, bringing American Sign Language translation to Pixel 11 smartphones through a new model called SL2T (sign-language-to-text).

The technology enables Deaf and hard of hearing users to sign directly to their phones anywhere they would normally type—whether searching the web, drafting messages, or interacting with Gemini. In Live Transcribe, users can sign responses in conversations rather than typing back and forth. According to Google DeepMind, testers found signing in ASL faster and more natural than typing in English.

Why it matters

While AI-powered speech recognition has become ubiquitous for hearing users, the estimated 70 million Deaf and hard of hearing people who use more than 200 sign languages worldwide have largely been excluded from this technological progress. This marks the first time sign language AI has moved from research labs into mainstream consumer products, potentially establishing a new baseline for digital accessibility.

Technical architecture addresses unique challenges

Sign language translation presents fundamentally different problems than speech transcription. Sign languages are independent natural languages with distinct grammars and lexicons, requiring true machine translation rather than simple word-for-word conversion. The model must also interpret simultaneous movements across hands, arms, torso, head, and face—a computationally intensive computer vision task.

SL2T processes sign language as pose landmark coordinates rather than raw video, protecting user privacy. An on-device MediaPipe Holistic model tracks body points, sending only geometric coordinates to servers for translation while immediately discarding the original video.

The model translates these coordinates directly into text, bypassing intermediate "glosses" commonly used in prior research. This approach removes artificial vocabulary constraints and allows translation quality to scale with available data.

Massive multilingual training

Google DeepMind trained SL2T on over 100,000 hours of data spanning more than 50 sign languages, with roughly 25% in ASL. Training jointly across diverse languages, dialects, and proficiency levels helps the model learn shared underlying structures, outperforming single-language models in internal experiments.

On the FLEURS-ASL benchmark, SL2T achieved a zero-shot score of 70 BLEURT—significantly higher than any previously reported result, according to Google DeepMind. The team also addressed practical deployment challenges including streaming latency, preventing hallucination on non-signing inputs, ensuring fairness for left-handed signers (10% of users), and improving one-handed signing performance.

Community-driven development

Google DeepMind established the AI Sign Language Advisory Committee (AISLAC), bringing together global Deaf organizations and experts to influence development priorities. The company co-authored a joint impact report with the committee detailing SL2T 1.0's capabilities and limitations for its Gboard and Live Transcribe release.

The feature is available at no additional cost on Pixel 11 devices, with expansion to more devices and additional sign languages planned. Details were first reported by Google DeepMind in an August 12, 2026 announcement.