Meta’s Muse AI agent can now talk. Mark Zuckerberg has announced real-time voice conversations for the assistant, along with what Meta calls expressive avatars, animated characters that move in sync with the agent’s speech. The update is rolling out now to Muse app users in the US, and Meta says it will reach its AI glasses soon.
Voice is no longer a differentiator in the AI assistant market. ChatGPT, Gemini and others have offered spoken, back-and-forth conversation for some time, so Muse is catching up on that front. What’s different is the face. Muse agents already had customisable avatars, and those avatars now lip-sync and animate while the agent speaks, so the conversation looks more like a video call with a character than a chat with a disembodied voice.
Users can also change how their agent sounds, either by simply asking it for a different tone and letting it choose, or by picking from a set of voice options in the avatar settings. Early impressions suggest replies come quickly over voice, with occasional pauses when the agent is looking something up, much as it does in text chat. The animation reportedly works well overall, although an avatar modelled on a real person can feel uncanny when its digital face starts talking back.
That uncanny effect is worth considering. Meta has made Muse highly personal, with an agent that has its own look and personality, and a talking face increases the emotional pull. That could make the product more engaging, but it also raises familiar questions about how attached people become to AI companions, especially when they come from a company whose business depends on attention.
So far, the momentum is real. Muse passed 10 million US downloads within weeks of launch, and Meta says roughly the same number of people have set up their own agents. Voice is an obvious next step to increase how often people come back to the app.
The bigger plan points beyond the phone. Meta says the voice and avatar features will also come to the Muse Charm, a dedicated hardware device the company says will ship this year. A talking agent makes more sense on a screenless wearable or a pocket device than in an app you already have to open and look at.
Availability outside the US hasn’t been confirmed, so users in the Middle East will have to wait. The more interesting question is whether people will want an AI agent with a talking face once the novelty wears off, or whether Meta is betting on a level of familiarity that many users may not be ready for.