Meta Rolls Out Muse Voice and Expressive Avatars in the US, Then Adds Custom Spaces

Meta on November 5, 2026 began rolling out voice conversations and expressive avatars in its Muse app in the US, on mobile, the web and Mac. A day later, on November 6, it added spaces, custom views that Muse builds from a description, and made them available everywhere Muse is available, the company said.

Both features build out Muse, the personal AI agent Meta launched in early September 2026, which Meta CEO Mark Zuckerberg said had 10 million users in the US and Canada at the time of the company's October earnings call (TechCrunch). Voice and avatars were first shown at Meta's Connect conference on September 23, and they reach the app about six weeks later as Meta's latest consumer AI product release, after the October 12 launch of its Muse Wonder models.

What did Meta launch on November 5?

Meta on November 5, 2026 started rolling out real time voice chat with an animated avatar in the Muse app, in the US to start, according to its announcement. People start a call by tapping the phone icon in the top right corner.

Meta says Muse keeps working on tasks in the background while the person talks, and that people can interrupt it mid answer. Users choose between voice only, the avatar next to the chat, or a full screen view, and pick from what Meta calls hundreds of voices, such as the energy of a New Yorker or the flair of a pirate. Meta pitches the mode for brainstorming, trip planning and rehearsing a job interview or a pitch.

The avatar reacts to what it hears with expressions and gestures, but Muse does not see the user, TechCrunch reports. Its default character is nicknamed Jolly, and users can change it to look like a person, an animal or another character. Meta says voice is coming soon to its AI glasses and to Muse Charm, a pocket device it announced at Connect.

How does Muse Realtime Avatar work?

Muse Realtime Avatar turns the speech stream of Meta's Muse Realtime Voice model into live video of a character, Meta's research team wrote in a technical post on September 23, 2026. Voice and video read the same stream of speech tokens, which Meta says keeps lip movement and expressions in sync with the audio.

The model is a diffusion transformer, a type of image and video generator that builds each frame by refining noise step by step. Meta says it cut the work per video chunk from 120 model passes to 2 through distillation, a method where a smaller student model learns to copy a larger one. It reports 448 by 768 pixel portrait video at 25 frames per second, about 870 milliseconds from the end of a user's turn to the first response, and 12 live sessions per Nvidia GB200 chip.

Meta also compared the system with two commercial avatar products. In its own rater study, people preferred Muse Realtime Avatar over Runway Characters 78% to 22% and over HeyGen LiveAvatar 88% to 12%, Meta says, though it adds that the mannerism comparison with Runway was not statistically different from a tie. Generated video carries an invisible watermark from Meta Video Seal.

What are spaces in the Muse app?

Spaces are custom views that Muse builds inside the app when a person describes an idea, such as a gift tracker, a word game or a planning board, Meta said in its November 6 post. People can keep refining a space with Muse, pin it to the sidebar or the main tabs, hide it, or schedule when it appears.

Spaces that draw on connected apps update themselves in the background, so price changes or new concert dates appear without a new request, the company says. Meta's examples include a shopping hub that pulls in Instagram saves and Marketplace listings, a school dashboard that finds events in school emails and adds them to a calendar, and a client hub for freelance work. The Verge lists Gmail, Google Calendar, OpenTable and Notion among the apps Muse connects to. Zuckerberg said in a post on X that he had built a BBQ game, an event tracker and a home control dashboard (TechCrunch).

Meta also says it is rolling out specialized versions of its models to Muse, trained with what it learned from Muse Wonder. The Verge reports these bring capabilities called Active Memory and Active Focus to Muse and the Meta AI chatbot.

How does Muse compare with OpenAI's dots?

OpenAI launched dots, its own always on personal agents with a cartoon persona, at its DevDay event on September 29, 2026, for ChatGPT Pro and Business Premium users, according to TechCrunch. Meta's voice and avatar mode, by contrast, is rolling out to all Muse users in the US.

The two products aim at different settings. TechCrunch reports that dots can be messaged through Slack, Teams and other workplace tools, while Meta's examples for Muse lean toward shopping, family logistics, travel and small business tasks. Zuckerberg said at Connect that Muse would be free for "a huge number of tokens" and that Meta expects to profit over time from small fees on transactions (TechCrunch).

What we don't know yet

• When voice and expressive avatars will reach Muse users outside the US.
• Whether voice calls or spaces carry usage limits or will cost money; neither announcement mentions pricing.
• The launch date and price of Muse Charm, and when voice reaches Meta's AI glasses.
• When people will be able to share spaces with others.
• How Muse Realtime Avatar performs in tests that Meta did not run itself.