Meta just gave its Muse AI agent a voice. CEO Mark Zuckerberg announced today that people can now have real-time voice conversations with their agents, and The Verge AI already has the feature, testing it on the iOS app while writing up the news. It’s rolling out now in the US, with Meta’s AI glasses and a new gadget next in line.
The early report is pretty good on speed. The weird part is the talking face, especially when that face belongs to Meta’s CEO.
What you need to know, in 6 points
- You can now talk to your agent live You speak, and Muse answers out loud in real time. It can still look things up or check something for you mid-conversation, the same way it does in a text chat.
- The avatar moves as it speaks The avatar at the top of your conversation animates in time with the agent’s words. The Verge AI says the animation works well, so this is more than a static profile picture with audio playing behind it.
- You’ve got two ways to change the voice You can ask your agent for the tone you want and let it decide. Or you can browse the options in the avatar settings. That first route is the more interesting one, because you’re describing a voice in plain language instead of scrolling through a menu.
- It’s fast, with some pauses In The Verge AI’s brief tests, Muse responded quickly over voice. It occasionally paused while looking up information or checking something, just as it would over text. Those gaps feel longer when you’re waiting on a voice, so they’re worth watching as more people get access.
- A talking Zuckerberg is as odd as it sounds The Verge AI’s reporter set the avatar to look like Mark Zuckerberg, as a constant reminder they’re chatting with a Meta service. Watching that digital copy talk back turned out to be “quite weird.” They also couldn’t get the voice to sound exactly like him. The report doesn’t say whether that’s a deliberate limit, but I wouldn’t be surprised if it is, given how touchy voice cloning has become.
- The hardware is where this is headed Meta says voice mode and expressive avatars are coming “soon” to its AI glasses. They’ll also be available on the forthcoming Muse Charm gadget, which is supposed to ship this year. And that’s the part I’d pay attention to, because glasses and a dedicated gadget make far more sense with voice than with typing.
Who can use it
- Rolling out now: the Muse app in the US
- Coming “soon”: Meta’s AI glasses
- Also planned: Muse Charm, expected this year
- Not covered: The Verge AI’s report doesn’t mention pricing or a timeline for other countries
Why it matters
Voice on its own won’t set Muse apart. ChatGPT and Gemini have offered voice modes for a while, so Meta is catching up there. The avatar is the more distinctive bet. Meta is wagering that people want to talk to a character with a face, and given its social roots, that’s not a surprising bet.
It’s also a risky one. A face that’s slightly off can bother people more than no face at all, and The Verge AI’s experience hints at that. The Zuckerberg avatar is an extreme test, though, since most people won’t pick a real person they recognize. Friendlier cartoon-style characters may not trigger the same reaction.
One caveat: this is a short hands-on from a single reviewer on iOS. We don’t yet know how the feature holds up in longer conversations or noisy places.
The next test comes when this reaches the glasses and Muse Charm, where voice stops being a nice extra and becomes the main way you use the product. More details are available in The Verge AI’s original report.