Previewed at Connect in September, Muse's talking avatar and customizable real-time voice began rolling out to every US user on November 5, 2026 — on mobile, web, and Mac.
2026-11-06 · by Moe Ameen
On November 5, 2026, Meta began rolling out real-time voice and a video-chat avatar to all US users of Muse, its personal AI agent from Meta Superintelligence Labs. The capability — branded Muse Voice — pairs a talking, animated avatar with two new models Meta calls Muse Realtime Voice and Muse Realtime Avatar. It was previewed at Meta Connect in September 2026; this week's rollout is the move from preview to general US availability, which chief AI officer Alexandr Wang marked on X with a demo of a dog avatar responding to his request for advice.
Mechanically, Meta describes a single shared stream of speech tokens that drives both the spoken reply and the avatar's video, so voice, lip movement, and expression stay in sync. You start a conversation by tapping a phone icon in the app, can interrupt mid-answer with a follow-up, and Muse can keep working a task in the background while you talk. There are three display modes — voice only, the avatar beside the chat, or the avatar full screen. The avatar is generated by the model rather than driven by your camera; its expressions and gestures are meant to match the conversation's tone, and you can set its appearance (human, animal, or another character) and pick from a large set of voices, changing the voice at any time. The feature is limited to users 18 and older.
Real-time conversations are live on mobile, the web, and Mac. Meta says voice for its AI glasses and for Muse Charm — a small wearable that shows the agent's avatar on a screen, planned before year-end — is coming later and is not part of this rollout. Per Meta's research write-up, video generated by the avatar model carries an invisible watermark via Meta Video Seal. Some finer specifications reported around the launch, such as exact resolution, frame rate, and response latency, come from single outlets, so treat those particular numbers as provisional and confirm them on Meta's own pages.
The headline most creators will take from this is "I can video-chat with an AI now." The more useful read is that two different jobs for an AI avatar just got pulled apart in public. Muse's avatar is a conversational interface — a face you speak to, that listens and answers. The avatar a creator actually grows an audience with is the opposite: one that faces outward, delivers your message, and lands as a finished, published video in the feed. [Kompozy](/) makes that second kind. Where Muse gives you someone to talk to, Kompozy generates [Persona Shorts](/glossary/persona-shorts) and [Persona Frames](/glossary/persona-frames) — HeyGen avatar video with one face held consistent across every clip, captioned, composited into brand-exact templates, and sized for each platform.
So act on the news without overclaiming it. Use Muse's voice mode the way it's good — think out loud, talk through an angle, let it draft a rough outline — then bring that into Kompozy as the seed for a content unit. Kompozy fans it into the formats a chat avatar can't produce: avatar and persona video, [Carousels](/glossary/hyperframes), Quote Graphics, Photo Posts, per-platform text, a blog, and a newsletter, all governed by a [Persona Brief](/glossary/persona-brief) so the voice is yours, then scheduled and published across eight social platforms plus blog and email with [Autopilot](/glossary/autopilot). For the agent's broader feature set see [Meta Muse](/ai-tools/meta-muse) and the [early-access wave](/news/meta-muse-early-access-advanced-features) this launch came out of.
Meta opened Muse Voice — real-time voice conversations plus a generated, talking video-chat avatar — to all US users. You can pick the avatar's appearance and voice, interrupt it mid-answer, and choose a voice-only, side-by-side, or full-screen view. It runs on two new models, Muse Realtime Voice and Muse Realtime Avatar, and was previewed at Meta Connect in September.
Not in the way creators usually mean. The Muse avatar is a face you video-chat with — a conversational interface for the agent — not a presenter that fronts finished, published videos. A consistent on-camera persona in Reels and Shorts your followers watch is generated by a content engine like Kompozy via Persona Shorts and Persona Frames.
At rollout it is live for US users on mobile, the web, and Mac, and limited to users 18 and older. Meta says voice for its AI glasses and the Muse Charm wearable device is coming later and was not part of this launch. Confirm current availability on Meta's own pages.
No. The avatar is generated by Meta's Muse Realtime Avatar model, which matches its expressions and gestures to the tone of the conversation rather than to a camera feed. Meta also says the generated avatar video carries an invisible watermark through Meta Video Seal.