Google DeepMind Launches SL2T: Sign-Language-to-Text Model on Pixel 11
dl_weekly · x · 2026-08-18
Google DeepMind has released SL2T, a massively multilingual sign-language-to-text translation model trained on over 100,000 hours across 50+ sign languages, achieving a zero-shot BLEURT score of 70. Integrated into Gboard and Live Transcribe on Pixel 11, it powers real-time ASL-to-English dictation. This allows Deaf users to sign for web searches, messaging, and interactions with Gemini, offering a faster and more natural communication method than typing.
More from Multimodal
- MiniMax H3 I2V struggles to maintain character physique across cuts — Acceptable-Thanks324 · 2026-08-18
- Seedance 2.5 Integrates into ContentStudio with 30s Generation and Character Consistency — SimplyAnnisa · 2026-08-18
- A 30sec One take Live Action Monster Film Created Using Seedance 2.5 — Suitable-Pin6964 · 2026-08-18
- Generating a superhero vs. Force battle video using MiniMax H3 — Devajyoti1231 · 2026-08-18
- AI-assisted musician pushes back on 'it's just Suno prompts' critics — Uncanny_Harry · 2026-08-18
- MiniMax H3 + USDU Workflow: 2K Video on 16GB VRAM — alisitskii · 2026-08-18