Google DeepMind launched SL2T, a sign-language-to-text model that powers sign-to-text dictation in Gboard and Live Transcribe on Pixel 11, starting with American Sign Language (ASL) to English. It is the first sign-language AI to reach a shipping consumer phone feature. It trained on more than 100,000 hours of signing across 50+ sign languages, about a quarter of it ASL, and it works from pose coordinates instead of raw video: an on-device MediaPipe Holistic tracker maps the signer’s landmarks, only those coordinates leave the phone, and the video is discarded. On FLEURS-ASL, a benchmark of studio-recorded ASL, SL2T scores 70 BLEURT, a translation-quality score, zero-shot, above every previously reported result.
A Deaf person can now sign to their phone anywhere they would normally type, to search the web, draft a message, or ask Gemini, and in Live Transcribe answer a conversation by signing instead of typing back and forth. Testers found signing in ASL faster and more natural than typing English. Google is careful about limits: a committee of Deaf organizations co-wrote a public impact report that restricts SL2T to casual, informal use, excluding medical, legal, and police settings. Known weak spots include rare signs, rapid fingerspelling, and low light.
The larger point is that sign languages just became a first-class input method on the world’s most common keyboard, at no extra cost. If the model extends beyond ASL, the roughly 70 million Deaf and hard of hearing people who use more than 200 sign languages get the dictation convenience hearing users already enjoy.
Read More: WAXAL expands speech data for 27 African languages
Sources:
- Putting sign language AI into users’ hands (Google DeepMind)
- Google DeepMind Brings Sign Language Translation to Phones With SL2T (Unite.AI)
- FLEURS-ASL dataset (Kaggle)
- AISLAC joint impact report for SL2T 1.0 (PDF)
Disclaimer: For information only. Accuracy or completeness not guaranteed. Illegal use prohibited. Not professional advice or solicitation. Read more: /terms-of-service
Reuse
Citation
@misc{kabui2026,
author = {{Kabui, Charles}},
title = {Gboard {Now} {Turns} {Sign} {Language} {Into} {Text}},
date = {2026-08-19},
url = {https://toknow.ai/posts/sl2t-sign-language-to-text-gboard-pixel-11/},
langid = {en-GB}
}