Meta has introduced Muse Voice, a new feature that brings real-time spoken conversations to its Muse AI assistant. Announced on 6 November 2026, it lets users talk directly with the assistant and watch an animated avatar respond as they speak — no typing, no waiting for a wall of text to appear on screen.

The feature is a fusion of voice interaction and Meta's Muse Realtime Avatar technology, which creates the facial expressions and gestures that accompany a conversation. Rather than asking a question and reading an answer, users interact with a customised digital character whose expressions shift with the tone and context of the discussion — different faces for different subjects, in the company's words, so the exchange feels more like talking to a digital partner than using software.

Starting a conversation is simple: users tap a phone icon in the interface, then choose how the assistant appears. There are three modes. Face-to-Face works like a video call with a personal AI avatar. A second mode places the avatar in the middle of a text conversation, so speech and text can be combined. And a voice-only mode lets users talk to the assistant without any animated character on screen at all.

One detail stands out for anyone who has used a voice assistant before: you can interrupt Muse mid-sentence. The assistant stops, listens, and lets you redirect the conversation or ask a new question without waiting for it to finish speaking — a small change that makes voice feel genuinely conversational rather than turn-based.

Customisation is central to the pitch. Meta says there are hundreds of voices and styles to choose from, so users can design how the assistant sounds and looks. That matters because Muse is positioned as a persistent digital identity, not just another chatbot window: it has long-term memory, retaining information across conversations over time to give more personalised answers, and agentic features designed to help with tasks rather than only answer questions. Unlike the Meta AI chatbot that appears inside services like WhatsApp, Muse has its own distinct look and interface.

Muse itself is still a young product — it launched in the US in September 2026 and has since rolled out in Canada. Muse Voice is rolling out to the United States first, on mobile, web and Mac, with more markets to follow; Meta has given no specific dates for international availability. The announcement also includes plans to bring Muse Voice to the upcoming Muse Charm device and to Meta's AI glasses, again with no launch dates attached.

The move fits a clear pattern: every major assistant is racing to become something you talk to rather than something you type at. OpenAI's voice mode, Google's Gemini Live and now Meta's Muse Voice all bet that the next interface for AI is spoken, expressive and interruptible. Whether users actually prefer chatting with an avatar over reading text will be the real test — but with a US rollout now under way, Meta is about to find out at scale.

For users, the practical question is what Muse Voice replaces. Meta is pitching it as the default way to use Muse, not a side feature: the assistant's long-term memory and agentic tools are meant to be reached by talking, whether that is planning a day, drafting a message or working through a task hands-free. The three modes matter here — face-to-face for a focused conversation, avatar-in-text for mixed use, voice-only for driving or walking — because they let the same assistant fit different moments instead of forcing one interface. The open question is international timing: with no dates given beyond the US rollout, users outside the US will be watching for the next announcement.

Sources