Meta has added spoken conversation to its AI assistant, the feature set it calls Muse. The new capability, Muse Voice, lets you talk to the assistant instead of typing.
Alexandr Wang, Meta's chief AI officer, announced it in a blog post and a post on X last Thursday.
What it actually does
During a conversation an animated avatar changes its facial expressions and moves its lips in sync with the speech. There are hundreds of voice options to choose from.
Meta says the feature runs on two components: Muse Realtime Voice and Muse Realtime Avatar. The avatar also reacts with facial expressions while you are speaking, not only while it is replying.
To start, you tap the phone icon at the top right of the conversation screen. You can interrupt the assistant mid-answer and ask something else — which Meta says makes conversations feel more natural.
Three display options are offered: voice only with no avatar, the avatar shown alongside the chat, or the avatar full screen.
The interruption is the real feature
Of everything announced, being able to cut in mid-response is the part that changes how the thing feels to use.
Voice assistants have historically worked in strict turns: you speak, it processes, it delivers a complete answer, and you wait through all of it even once you know the answer is wrong. Real conversation does not work that way — people interrupt constantly, and that is how they steer a discussion without losing time.
Supporting it means the system has to keep listening while it is talking, and be able to abandon a response part-way. That is a harder engineering problem than generating the speech, and it is the thing most likely to make people use voice rather than typing.
The avatar is a choice worth noticing
The lip-synced face reacting to you while you talk is doing something other than conveying information. It is signalling attention — the visual equivalent of a nod.
That is effective, and it is also the oldest trick in interface design: a system that appears to be listening is trusted more than one that simply is. Giving users the option to switch the avatar off entirely, which Meta has done, is the right call, because the voice-only mode is the honest version of the same product.
Where this leaves readers here
It launched for all users in the United States, on Meta's mobile app, its website and Mac computers, rolling out in phases. It is not available in other countries, and Meta has given no timeline.
That pattern is now familiar and worth planning around rather than waiting on. Voice features depend on speech models trained per language, and Bangla has nothing like the training data English does — so Bangladesh tends to sit near the end of these rollouts even after the geographic restriction lifts.
If you see Muse Voice working here before an official launch, it is almost certainly a VPN placing you in the US, which puts your Meta account at risk for a feature that will arrive properly on its own.
