Google brings sign-language-to-text to Pixel 11 with SL2T model
Google DeepMind has integrated a sign-language-to-text model called SL2T into Pixel 11 devices, enabling users to sign in place of typing. The feature is available through Gboard and Live Transcribe, beginning with American Sign Language to English translation. Google plans to extend support to more devices and sign languages.
Editor, Lazyfounder

TABLE OF CONTENTS
KEY HIGHLIGHTS
- Google DeepMind introduced SL2T, a sign-language-to-text model for Pixel 11 devices.
- SL2T powers sign-to-text features in Gboard and Live Transcribe, starting with American Sign Language to English.
- The model was trained on over 100,000 hours of data across more than 50 sign languages.
- On-device processing ensures privacy by using body pose landmarks instead of raw video.
- Google plans to expand device support and add more sign languages over time.
How the feature works
SL2T translates sign language into written text by analyzing body movements, including hands, arms, torso, head and face. Unlike speech-to-text, which converts spoken words into written form, sign-language-to-text requires machine translation to interpret visual language and produce fluent written output.
The model was trained on over 100,000 hours of data spanning more than 50 sign languages, with about a quarter of the dataset focused on American Sign Language. This broad training helps the system recognize patterns across languages, dialects and signing styles.
Privacy and on-device processing
Google designed SL2T to run on-device to address privacy concerns. Instead of sending raw video footage, the system tracks body pose landmarks—geometric points representing movement. These coordinates are sent for translation, while the original video is discarded immediately.
This approach avoids the need for intermediate labels, often called glosses, which can miss nuances like facial expressions, spatial meaning and non-manual markers. By translating movement coordinates directly into text, the model aims to capture more of the language’s complexity.
Real-world challenges and user involvement
Google reports that SL2T performs well on ASL-to-English benchmarks but acknowledges limitations in real-world use. The team focused on reducing delay, supporting left-handed signers, handling one-handed signing and avoiding incorrect outputs when no signing is detected.
Deaf users and experts were involved throughout the development process, including testing, data collection and impact assessment. Google also formed an AI Sign Language Advisory Committee to guide the project.
Availability and next steps
The feature is rolling out on Pixel 11 devices through Gboard and Live Transcribe at no additional cost. Google says it will expand support to more devices and sign languages over time.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
This launch signals two shifts for founders and operators. First, it shows that AI-driven accessibility tools are moving from research prototypes to mainstream consumer devices, which may accelerate adoption and raise user expectations. Second, the use of on-device processing for privacy-sensitive visual data could influence how startups design AI features for regulated or high-trust markets, particularly in healthcare, education and enterprise.
Key takeaways
- Founders in accessibility tech can see a concrete example of AI moving from research to consumer products, with implications for product design and user interaction.
- The shift to on-device processing for privacy-sensitive applications sets a precedent for how AI models may handle visual data in regulated markets.
- Support for multiple sign languages signals a growing market for localization in accessibility tools, particularly in regions with underserved linguistic communities.
- Involving Deaf users and experts in development highlights the importance of community collaboration for products serving marginalized groups.
Sources
- yourstory.com
Google puts AI sign language translation on Pixel 11
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.


