Meta announced on November 5, 2026 that voice conversations and expressive avatars are rolling out in its Muse app in the United States. Its instructions cover Muse on mobile, the web and Mac: tap the phone icon at the top right of the conversation to start talking. The release is a rollout, so it should not be read as a promise that every account or country already has the feature. Meta says voice on AI Glasses and Muse Charm is still coming later.
What can you do with Muse Voice now?
Meta describes a live conversation with its personal AI agent rather than a voice recording sent for a later text reply. Users can speak, interrupt a response and choose from voice styles. They can keep the interaction voice only, place an animated avatar beside the chat, or use a full-screen avatar view. The announcement suggests brainstorming, planning and rehearsing as examples. Those are Meta's proposed uses, not tasks VIBPS has tested or measured.
The distinction between a live conversation and an assistant making an outbound phone call matters. This release concerns talking with Muse inside Meta's app. It does not establish that Muse can call any person or business on the user's behalf, and it does not announce an open developer API for the new avatar. A maker evaluating the experience should check the actual app and account availability rather than assuming the research demo is a product integration option.
How does the avatar follow the conversation?
Meta's earlier research report explains the software connection. Muse Realtime Voice produces a stream of speech tokens carrying both the words and their delivery. A speech decoder and Muse Realtime Avatar use that stream so the audio, lip movement and expression can stay synchronized. The avatar model also uses reference media and recent generated frames to maintain a consistent character across a longer conversation. This is Meta's account of its architecture, not an independently reproduced benchmark.
Meta reports a 448-by-768 portrait stream at 25 frames per second and about 870 milliseconds from the end of a user's turn to the first byte of synchronized output in its research setup. That number is not a measured latency for every phone, connection or Muse session. The research examples also include characters that may not all appear in the consumer app; Meta explicitly warns against assuming every demonstration avatar is available to users.
Who can use Muse's new voice and avatar features?
Meta says the current rollout starts in the US on mobile, web and Mac. The company has not given a date for a wider geographic launch or for voice conversations on its glasses and Charm device. Its research page says Muse is for people aged 18 or older. Readers outside the US should treat videos from the launch as a preview of a possible future experience, not evidence that their own account is eligible today.
For app makers, the release is an example of a broader interface shift: a software experience can combine conversation, timing and a visual character, with each part needing to feel coherent to the person using it. That observation does not establish that avatars improve task completion or user trust. VIBPS is a social network for software where people can follow makers and their apps; it did not build Muse and has no partnership with Meta. The useful question is what this particular app now makes available to try.