Meta Rolls Out Real-Time Voice and Expressive Avatars for Muse AI Agent in the US

Meta has launched Muse Voice, giving users in the US real-time voice conversations with the company’s consumer AI agent. The feature supports both calls and video chats, with Muse responding aloud while an animated avatar moves in sync with its speech.

The rollout brings Muse beyond text-based requests and into a more social interface. Users can ask questions, request tasks, and continue a conversation while the agent searches for information or performs work. The feature is available on mobile, web, and Mac, with Meta AI glasses and the planned Muse Charm device next in line.

Muse’s defining feature is its customizable animated character, which Meta designed to make conversations feel more personal. Users can choose an appearance resembling a person, animal, or other character, then adjust the voice and tone through settings or by asking Muse to make a choice.

The avatar uses subtle expressions and gestures that match the conversation. It doesn’t see the user, but it hears spoken requests and reacts to the exchange. A fluffy white dog avatar appeared in a demonstration by Meta chief AI officer Alexandr Wang, while a Verge test used a digital likeness of Mark Zuckerberg. The latter produced an intentionally strange result, showing how quickly expressive avatars can move from friendly to uncanny.

Initial hands-on testing found that Muse responds quickly over voice, though pauses occur when it checks information or completes a request. That behavior mirrors text interactions, where the agent can take time to retrieve an answer or carry out a task.

The animated presentation worked technically, but the effect varied by avatar. Expressions and mouth movements helped the character feel active, while the Zuckerberg likeness made the interaction feel unsettling. Voice customization also has limits: the test could not reproduce Zuckerberg’s voice exactly, underscoring the distance between a recognizable visual persona and a fully convincing digital replica.

The voice rollout arrives before Meta’s planned release of Muse Charm, a Tamagotchi-like device expected before the end of 2026. The small wearable will display the Muse avatar and let users speak with it directly, turning an AI assistant into a persistent companion rather than an app opened only for specific requests.

Meta is also preparing the feature for its AI glasses. That sequence connects the same agent across screens, phones, glasses, and dedicated hardware. The strategy gives voice interaction a role in Meta’s wider consumer AI push, while the avatar supplies a recognizable identity across products.

Meta is betting that a face, voice, and responsive gestures will make people feel more connected to Muse. That design choice can make an assistant easier to approach, especially for users who find spoken conversation more natural than typing. It also gives ordinary requests a stronger sense of social presence.

The tradeoff is that anthropomorphic design can make an automated system feel more capable or emotionally aware than it is. Muse’s reactions come from its conversational models, and the agent does not visually perceive the person speaking to it. As Meta expands Muse into glasses and a dedicated wearable, the quality of those boundaries, along with the consistency of voice and avatar behavior, will shape whether the feature feels useful or merely theatrical.

Muse Voice is rolling out now to US users, while support for Meta’s AI glasses and Muse Charm is scheduled for a later release. The immediate test for Meta is adoption: whether people return for spoken conversations after trying the animated interface.

The next phase will also expose practical limits. Users will judge response latency, information accuracy, voice variety, avatar customization, and how comfortable the agent feels in public settings such as glasses. Meta’s rollout places Muse in direct competition with other AI assistants that already offer voice chat, but its combination of live speech, expressive characters, and planned hardware gives the product a distinct consumer pitch.