Muse voice: how to call and video chat with your Muse

Muse voice lets you hold a live spoken conversation with Meta's Muse agent, either audio only or with an animated avatar on a video-style call. It's available in the US in the Muse app on mobile, web and Mac, and you start a call by tapping the phone icon in the top-right corner of a conversation.

What's been posted

@TheFederal_News described the rollout as "a new video calling feature" that lets people talk face to face with a customized AI agent. Meta's November 5 announcement has the details, including the claim I care most about: "Muse handles it in the background as you talk."

Meta has been making that promise since Connect in September, when @finkd wrote that Muse will "work in the background and get things done for you" during a conversation. Now people in the US can test it.

One naming trap: a "call" here means you talking to your own Muse inside the app. It's different from the outbound calling beta, where Muse rings a business for you.

How to use Muse voice

1. Open Muse on your phone, on the web or on your Mac, and go to a conversation. If you don't see the phone icon, update the app first.
2. Tap the phone icon in the top-right corner.
3. Pick a view: voice only, the avatar next to your conversation, or full screen.
4. Start talking. You can cut in while Muse is speaking.
5. Try a different voice if the default grates on you. Meta says there are hundreds and you can switch whenever you want, though its post doesn't walk through the menu.

To see whether the background work is real, give it a small job in the middle of a conversation. These prompts are mine, so swap in your own details: "While we talk, draft a short email telling my Tuesday customers we're running a day behind." Change the subject for a minute, then come back and ask, "Read me that draft, and don't send it." Interrupt halfway through and say, "Make it shorter." If it keeps up, you've learned more than the launch post can tell you.

What Muse voice can do

The response-time figure comes from Meta's September research post. It's Meta's own number, so don't expect it on hotel Wi-Fi.

What Meta hasn't said

The announcement doesn't mention price, plan limits, supported languages or how voice conversations are stored. It describes what Muse says and shows, with nothing about whether it can use your camera. If you plan to talk through anything involving customers or staff, check your privacy settings first.

Take the quality claims with some salt too. In that research post, raters preferred Meta's avatar 78% to 22% over Runway Characters and 88% to 12% over HeyGen LiveAvatar after two- to three-minute conversations. Meta was grading its own system, and the post doesn't say how many people took part. It also concedes that the comparison with Runway on mannerisms was statistically a tie, and that its examples "do not all reflect avatars available in the Muse app."

I think the avatar is the least important part of this. A face makes a good demo, but the useful claim is the one about background work, and Meta offers no numbers for how often those tasks finish correctly.

What this means for businesses

Voice suits people whose hands are busy, and that describes a lot of trade shop and restaurant owners. You could dictate tomorrow's plan from the truck and have the draft waiting when you park.

The catch is that spoken instructions are quick to give and hard to review. Muse's Active Memory update makes that matter more, since something misheard on a call could stick around. Keep the habit you'd use in text and have it read back anything that leaves the building. Meta's post doesn't say whether the Sentinel approval check works any differently during a voice conversation, and I wouldn't assume it does.

If you're trying to work out which parts of your week could safely run on a spoken request, New Face Design offers a free process audit.