Meta's Muse AI agent is getting a voice. A new real-time voice chat mode lets users talk to the agent conversationally — and have it carry out work while they, as Android Authority puts it, sit back and relax. The mode ships with hundreds of voices to choose from.

The update moves Muse from a text-driven tool toward something closer to a spoken delegation layer: describe a task out loud, let the agent execute. That framing matters because the industry's agent race is increasingly about reducing friction between intent and action, and voice is the lowest-friction input there is.

The sheer size of the voice library is its own signal. Hundreds of options goes well beyond the handful of personas most assistants offer, suggesting Meta sees voice personalization as a retention hook — the same bet behind its celebrity-voiced AI characters. The move also sharpens competition with OpenAI's voice mode and Google's Gemini Live, both of which pair conversational speech with task execution. For Meta, embedding a capable talking agent across its apps is the distribution advantage the others can't easily match.