Meta AI research breakthrough enabling ultra-low-latency real-time video avatar calls from voice chats.
Published by Meta AI Research with full scientific papers and open demonstration samples.
Muse Realtime Avatar is a major computer vision and multimodal milestone from Meta AI Research. Moving past traditional avatar pipelines that demand heavy offline rendering, Muse employs an end-to-end neural rendering architecture capable of producing synchronized facial expressions, precise lip motions, and expressive head gestures with sub-150ms glass-to-glass latency. This bridges conversational voice agents into real-time interactive face-to-face video telepresence.