Meta Muse Voice Is Live, and Spaces Turns Your Chats Into Mini Apps

Meta Muse Voice Is Live, and Spaces Turns Your Chats Into Mini Apps

Meta gave its Muse assistant a voice this week, then gave it a workshop the next day. On November 5, real-time voice conversations and expressive avatars rolled out in the Muse app in the US. On November 6, Spaces arrived: custom mini-apps you build inside Muse by describing an idea. The same day, Meta rolled out better memory and focus to the models behind Muse. It is the biggest two-day overhaul Muse has had since launch, and the most interesting parts are the ones that are not available yet.

What Muse voice actually does

Tap the phone icon in the top-right of the Muse app and you can talk to Muse in real time. The voice answers right away, and you can interrupt it mid-sentence, the way you would interrupt a person. An animated avatar sits at the top of the conversation, moving its expressions and gestures with the tone of what is being said, and reacting while you talk.

The avatar is optional. Meta lets you pick voice only, the avatar alongside your conversation, or full screen when you want to talk face to face. There are hundreds of voices to choose from, and you can change them whenever you want. Muse can also carry out tasks in the background while you are still talking.

Two systems sit underneath, and Meta names both: Muse Realtime Voice and Muse Realtime Avatar. Meta previewed the feature at its Connect conference in September, and the research behind the avatar was published the same day. In blind live trials, Meta says its avatar beat Runway 78 percent to 22 percent and HeyGen 88 percent to 12 percent. Those are Meta's own numbers, and nobody independent has verified them.

The catch is geography. Voice and avatars start in the US only, on mobile, web, and Mac. Muse itself is now in five countries, and Mexico launched on November 5, but Meta has given no date for voice outside the US. Voice on Meta's AI Glasses and the Muse Charm wearable is coming soon, with no dates attached.

Spaces: describe it, get an app

Spaces, launched November 6, is the more ambitious half of the update. Describe an idea in plain chat and Muse builds a custom interactive screen inside the app: a holiday gift tracker, a daily word game, a planning board, a school dashboard, or an order tracker for a small business. Meta lists nine example spaces, seven personal and two for small business or freelance work, including a wedding planner and a client hub.

Spaces stay updated in the background from apps you connect. The Verge lists Gmail, Google Calendar, OpenTable, and Notion among the connectors. Spaces are available everywhere Muse is available, starting November 6. Sharing a space with other people is not possible yet. Meta says it is exploring that, which is the honest way of saying it does not exist yet.

This is the part of the launch that could matter most, because it turns Muse from a chatbot into a place where small tools live. A gift tracker that updates itself from your email is genuinely useful. A chatbot that can also hold a conversation is not a new idea, but a chatbot that builds you a working dashboard from one sentence is a different product.

The memory upgrade underneath

The same day Spaces launched, Meta rolled out specialized models trained using its Muse Wonder series, with Active Memory and Active Focus. Meta's own measurements claim a 2.4x improvement in one memory metric and a 67 percent improvement in another. Again, these are Meta's figures, unverified by anyone independent, so treat them as a direction rather than a score.

What matters practically is whether Muse remembers what you told it last week without being reminded, and whether it stays on task during a long session instead of drifting. Those are the two things that make an assistant feel like an assistant instead of a search box with a voice. We have not tested Muse Voice, Spaces, or Active Memory hands-on, and nobody outside Meta has either, so the honest verdict is: promising on paper, unproven in practice.

What is still missing

The gaps are easy to list. Voice is US-only with no timeline for anyone else. The avatar cannot be designed by the user, as far as Meta's announcements say, and Meta has not said whether you can pick a human-looking one. Spaces cannot be shared. Meta has not said what happens to voice recordings, or whether they have anything to do with ads. And the glasses and Charm versions of voice have no dates.

That list matters because Meta is clearly building Muse into a platform, not a feature. The open-weight Muse Spark model showed Meta wants the model itself to travel, and the Muse Wonder series showed it wants the reasoning to get cheaper. Voice, Spaces, and memory are the consumer-facing half of the same plan: make Muse the place you talk to, build in, and come back to.

Should you try it

If you are in the US, try the voice first. It costs nothing, it is one tap, and you will know within a minute whether talking to an assistant feels better than typing to one. Try Spaces second, with something small and concrete, like a tracker for a trip you are actually planning. Skip the memory claims until you have used Muse for a week and can judge for yourself whether it remembers anything.

If you are outside the US, there is nothing to do but wait. Meta has not said when voice reaches other countries, and the Spaces launch is the only part of this update you can use today. That split, US voice now and everyone else later, is the oldest pattern in Meta's playbook. The product is real. The rollout is not finished.