Following its November 5 US rollout, Meta's Muse live avatar calls are available on mobile, web, and Mac, not only the Muse app.
The calls also support interruptible voice conversations with the avatar during sessions.
First citedThe Neuron
FollowingDay 3in the edition since 6 Nov 20263 sourcesupdated 7 Nov 2026
Meta Muse AI avatar live video calls
The gist
Latest: Meta's Muse AI Avatar Opens Live Video Calls to All US Users
Meta is making real-time AI avatar video calls available at consumer scale, with technical tradeoffs around hardware costs: serving 1 million simultaneous callers would require roughly 83,000 GB200 GPUs for video alone, which could cap the avatar's reach.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Meta has made live video calls with its Muse AI agent available to all US users across mobile, web, and Mac. The feature pairs an animated avatar that reacts in real time with interruptible voice conversations. Muse Realtime Avatar, previewed at Meta Connect before its public rollout, converts Muse Realtime Voice into expressive, interactive avatars. The system uses two specialized real-time models: Muse Realtime Voice for speech token generation and Muse Realtime Avatar for lip-synced video rendering. Meta distilled a student model that reproduces in 2 passes per video chunk what a teacher model required 120 passes to produce, enabling real-time performance. The avatar streams 448x768 portrait video at 25 fps with approximately 870ms end-of-turn latency. A single Nvidia GB200 can handle 12 simultaneous video streams using 4-bit quantization, caching, and dynamic batching. In internal tests, human raters preferred Meta's avatar over Runway Characters 78% of the time and over HeyGen LiveAvatar 88% of the time.
How it developed
First in the edition · 6 November 2026
Meta launched live video calls with its Muse AI avatar to all US users on November 5, pairing two real-time models: Muse Realtime Voice for speech generation and Muse Realtime Avatar for lip-synced video rendering at 448x768 portrait resolution and 25 fps.
To hit real-time performance, Meta distilled a student model that reproduces in 2 passes what a teacher model required 120 passes to produce, with a single Nvidia GB200 handling 12 simultaneous streams. In internal tests, human raters preferred the avatar over Runway Characters 78% of the time and over HeyGen LiveAvatar 88% of the time.
Sources
- Google DeepMind Launches Gemini 3.8 Live with Real-Time Video Avatars
- Meta Superintelligence Labs Muse Voice Transcribe launch
- Meta's Muse Glimmer 30B Runs AI Agents on a Single Consumer GPU
- Meta's Muse Code AI coding agent
BackgroundAI agents: models that take actions
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.