Meta unveils real-time avatars that speak and move with voice
Meta introduced Muse Realtime Avatar, which turns Muse Realtime Voice speech tokens into expressive live video of characters, portraits, animals or objects. Meta says it streams 448x768 portrait video at 25 frames per second with about 870 ms latency, keeping voice, lip motion and expression synchronized. In live-call tests against Runway Characters and HeyGen LiveAvatar, Meta says raters preferred its system overall, and videos carry an invisible Video Seal watermark.