Meta's new speech model targets glasses and agents in messy real-world rooms
heypearlai · x · 2026-09-04
Meta's newly released speech model is positioned as the 'ears' for its glasses and agent stack, built to handle messy real rooms instead of waiting for a clean 'Hey Meta' command.
- Beyond the error rate, the notable part is the intended use: voice entry for Reality Labs glasses and the agent stack
- It is live now via the Meta Model API, Meta AI for Mac, and Muse Code
- The author questions whether it holds up outside demos recorded with 20 consenting participants
More from Embodied
- Open-source Microduck teaches robots to play football, easing entry to autonomous robotics — tristanbob · 2026-09-04
- Tesla opens Cybercab fleet purchase interest form, signaling third-party robotaxi fleets — JOBhakdi · 2026-09-04
- Open source artificial muscles are coming sooner than you think — IanPritchard · 2026-09-04
- Palmer Luckey hypes mystery silver headset Dime, as investigation digs into OpenAI hardware tease — rohanpaul_ai · 2026-09-04
- $14k Chinese Cars Now Ship With LiDAR, No Subscription — and BYD Covers Crash Losses — aronchick · 2026-09-04
- Losing a badminton match to a robot: a surprising 2026 milestone — lukas_m_ziegler · 2026-09-04