Meta’s Muse AI agent adds real-time voice and expressive avatars

Meta has rolled out real-time voice and animated avatars for its Muse AI agent in the US, giving creators and social operators a new way to test interactive AI personas.

What happened

Meta announced on November 5, 2026 that its Muse AI agent now supports real-time voice conversations in the US. CEO Mark Zuckerberg shared the update, and The Verge’s Jay Peters tested the feature on the iOS app. During voice chats, the Muse avatar at the top of the conversation animates in time with the words the agent speaks.

Users can change the voice tone by asking the agent in natural language to adjust it, or by browsing options in the avatar settings. In brief tests, Peters found the agent responded quickly over voice, though it occasionally paused while looking up information or checking something, similar to how it would behave with a text chat query. The voice mode and expressive avatars are rolling out now in the US for the app, with support coming “soon” for Meta’s AI glasses and the forthcoming Muse Charm gadget, which is supposed to ship this year.

Key facts

• Meta’s Muse AI agent now offers real-time voice conversations, announced by CEO Mark Zuckerberg.
• The feature is rolling out now in the US for the app; The Verge tested it on iOS.
• Muse avatars animate in sync with spoken words, and voice tone can be changed by asking the agent or via avatar settings.
• In brief tests, Muse responded quickly over voice but occasionally paused while looking up information or checking something.
• Support is coming “soon” for Meta’s AI glasses and the forthcoming Muse Charm gadget, which is supposed to ship this year.

Our analysis

The shift from text responses to real-time voice and synchronized animation changes what Meta’s Muse agent can be used for. It moves the tool closer to an interactive character than a chat box. For creators and social media managers, this likely makes Muse more relevant for producing audio-first or avatar-driven content, and for testing how audiences respond to a persistent branded persona.

One operational detail matters: the agent is fast but not instant. Peters noted occasional pauses while the system looked up information or checked something, suggesting that operators should not yet treat Muse voice as a zero-latency live host. The “uncanny” quality of the animated Zuckerberg avatar also suggests that avatar realism is uneven; brands considering a digital spokesperson should test audience comfort before assuming the format feels natural.

The announced expansion to Meta’s AI glasses and the Muse Charm gadget indicates Meta wants voice and expressive avatars to live beyond the phone screen. That likely means more ambient, hands-free touchpoints where a creator’s or brand’s Muse persona could interact with users in short conversational moments.

What it means for operators

• Test Muse voice mode on iOS in the US and record response speed, pause patterns, and the exact voice-tone prompts that produce the desired result.
• Write short conversational scripts for Muse because pauses may occur while it looks up information; concise prompts can keep interactions flowing.
• Adjust voice tone by asking the agent in natural language or via avatar settings, and document the setup to keep a consistent brand or creator voice.
• Track the AI glasses and Muse Charm rollout, and prepare audio-first or hands-free interaction plans if your audience uses Meta wearables.