Meta’s Muse app now talks back with a live avatar, US users first

Meta has begun rolling out voice conversations and animated avatars in its Muse app. The company said in a post on X that Muse Realtime Avatar, previewed at Connect, is now live in the app in the US.

Talk to Muse, and watch it respond

Meta’s newsroom post describes two linked pieces. Muse Realtime Voice handles the spoken conversation. Muse Realtime Avatar adds an animated character that reacts during the exchange, with expressions and gestures matched to the tone of the talk.

Meta says users start a conversation by tapping the phone icon in the top-right corner of the app. They can interrupt Muse while it is answering, and Muse keeps working on tasks in the background while the conversation goes on. There are three display modes: voice only, the avatar next to the conversation, or the avatar full screen. Meta also lists hundreds of voice styles that can be changed at any time.

The rollout covers the Muse app on mobile, web and Mac, in the US to start. Meta’s own line is “Voice means Muse goes with you”, and it says voice conversations are coming soon to its AI glasses and to Muse Charm. The post gives no price, no age requirement and no date for other countries.

What Meta has said about the technology

At Connect, Meta framed the avatar as real-time embodiment technology: it takes the output of the voice model and turns it into an interactive character. The newsroom post links to a separate Meta research post for the technical side, which was not available in full to us.

Coverage of the September announcement, including RuntimeWire, reported figures that Meta supplied: about 870 milliseconds from the end of a user’s turn to the first byte of a synchronized voice-and-video reply, and 12 video sessions per GB200 accelerator. The same report says generated output is watermarked and that Meta’s preference comparison was run by the company itself.

Limits and open questions around a US-only launch

The announcement leaves several things open. Meta has not said when the feature reaches other countries, and the post does not describe any safety or privacy measures for talking to a lifelike character, such as how voice recordings or conversation data are handled. The 870 ms latency and the sessions-per-GPU figure come from Meta and have not been independently checked, and the research post reportedly cautions that its demos do not all show avatars available in the app today.

Context helps here. According to TechCrunch, Muse launched in early September 2026 on top of Meta’s Muse Spark model, and Meta said then that the app is free for “a huge number of tokens”, with the company expecting to earn a small fee from transactions such as shopping. That matters for the avatar: live video generation is costly to serve, and the 12-sessions-per-GB200 figure suggests why a staged, US-first release makes sense. TechCrunch also reported that Meta received more than 1,500 applications for developer-built connectors in under a week, a sign of how fast the company is trying to widen what Muse can do.

The avatar also sits in a crowded space of voice assistants. Nearby OpenAI features, such as the image tools covered in our report on ChatGPT’s new GIF support, show the same race to make assistants more visual. Meta’s difference is the combination of an always-on voice, a face, and delegated tasks in one app.

For readers in the US, the practical step is simple: update the Muse app and look for the phone icon. Everyone else has to wait, and Meta has given no timeline. Whether people prefer talking to an animated character over plain voice is untested outside Meta’s own comparisons, and the display modes suggest the company expects some will not.