Meta is giving its Muse AI agent video chat, although Muse cannot see the person on the other end. The new Muse Voice feature pairs real-time conversations with an animated, customizable avatar, giving users a character to talk to as they ask questions or request tasks. The U.S. launch also sets up the experience planned for Meta's AI glasses and its forthcoming Muse Charm wearable.
What video chat means here
The visible participant is the avatar. As Muse speaks, its character animates along with the words, using expressions and gestures intended to suit the conversation's tone.
Muse hears the user and responds to what they say. It does not see them. That distinction matters because the video-chat label can suggest a two-way visual conversation, while the feature described here gives the AI a visible presence without giving it a view of the person speaking.
Meta says the experience uses new models and technology called Muse Realtime Avatar and Muse Realtime Voice. The feature was first introduced at the company's Connect developer conference in September, before this November 5 announcement of U.S. availability.
The underlying interaction remains familiar: ask a question, have a conversation or request that the agent handle something. The addition is a speaking character whose movements accompany its replies. There is now a face to watch while the AI does its thinking.
Users can change the character and its voice
Muse's animated character is nicknamed Jolly, but users can customize the avatar to resemble a person, an animal or another character. Meta is leaving room for people who would prefer their assistant to be a fluffy dog, as well as those with more particular casting requirements.
Voice customization works in two ways:
- Ask Muse directly. Users can describe the tone they want and let the agent choose how to adjust its voice.
- Choose through settings. Users can browse voice options in the avatar settings instead.
Those controls allow users to shape how the agent sounds, but a chosen appearance does not establish that its voice can reproduce the same person's sound.
In one brief iOS test, the avatar had been customized to resemble Mark Zuckerberg. The tester could not get its voice to sound exactly like him. A recognizable face and an exact vocal match were separate achievements, and only the first had been managed.
The U.S. rollout has an availability ambiguity
CEO Mark Zuckerberg announced the addition of real-time voice conversations. Meta also described the launch in a blog post, while chief AI officer Alexandr Wang shared a demonstration on X.
The company says real-time conversations are supported on mobile, web and Mac. A brief hands-on test confirms that the feature had reached at least one user of the iOS app.
The two launch reports differ in how they describe availability. One says Muse Voice is now available to everyone in the U.S., while the other says voice mode and expressive avatars are rolling out now in the U.S. app.
That leaves an ambiguity about whether every U.S. user should already have access. Both accounts place the launch in the U.S., but they do not describe its completion in the same terms. Neither account establishes availability outside the country.
An early test found quick replies, with occasional pauses
The brief iOS test found that Muse generally responded quickly over voice. It sometimes paused while looking up information or checking something, much as it would during a text conversation.
That is a useful distinction for anyone expecting a continuously flowing exchange. A spoken interface can make the interaction feel more conversational, but the agent may still need a moment to complete the work behind a reply.
The avatar's animation also tracked its speech well in that test. The result was nevertheless described as uncanny, particularly because the chosen character resembled Zuckerberg.
These are early impressions from limited use, not a broad evaluation of response speed or animation quality. They do, however, separate two things that a polished demonstration can blur together: whether the agent replies promptly, and whether watching it speak feels comfortable.
For someone trying the feature, those are sensible things to judge separately. A preferred voice may make an exchange more pleasant, while pauses during information requests remain part of how the agent works.
Meta wants a character people feel connected to
Other AI products already offer voice conversations. Meta's stated rationale for adding an animated character is that making Muse more humanlike could help consumers feel more connected to their agent. Live conversations could, the company believes, deepen that connection.
The expressions and gestures serve that goal. Their purpose is to make the avatar react in a way that fits the exchange, rather than simply move its mouth through an answer.
Wang's demonstration offered a casual example. His Muse appeared as a fluffy white dog while he joked about his fondness for posting memes on X and suggested that some posts might have caused trouble with his boss.
The avatar responded by suggesting disclaimers for memes that could otherwise be mistaken for Meta endorsing provocative material. Even an animated dog, apparently, can find its way into a discussion of corporate communications.
The demonstration showed the conversational style Meta is promoting. It did not establish whether an expressive avatar actually makes users feel closer to an AI agent. That remains the company's expectation, separate from the announced ability to speak with one.
Muse Charm gives the avatar somewhere else to appear
Meta plans to bring real-time conversations to its AI glasses and the forthcoming Muse Charm. Support for the glasses is described as coming soon, without a specific date in the reports.
Muse Charm is a planned Tamagotchi-like wearable with a small screen on which the avatar will appear. Users will be able to speak to the agent, which can reply or carry out tasks on their behalf.
The device is expected to ship before the end of 2026. That is a shipping target, not an announcement that the hardware is available now.
The software launch therefore introduces the voice-and-character interaction before the dedicated device arrives. Users can begin talking to an animated Muse on existing platforms while Meta prepares another place for that character to live.
For now, the voice is arriving in software. The wearable still has to ship.