正文 · AI 翻译
IT Home reported on November 6 that Meta CEO Mark Zuckerberg posted a thread on the social platform Threads today (November 6), announcing that the personal AI assistant Muse has officially launched real-time voice conversations and expressive avatar features.
The features are rolling out first to users in the United States, across mobile, the web, and the Mac app. By tapping the phone icon in the upper-right corner of the app, users can talk to Muse as if making a phone call and interrupt it at any point while it is responding. IT Home has included relevant screenshots below:
On the technical side, the update is powered primarily by the Muse Realtime Voice speech model and the Muse Realtime Avatar model. The former generates information about both “what to say” and “how to say it” simultaneously; the latter uses the same data stream to generate facial expressions, lip movements, and body movements in real time, keeping the audio and visuals in sync.
Meta says the system can stream vertical video at a resolution of 448×768 and 25 frames per second, with a delay of about 870 milliseconds between the user finishing speaking and receiving the first audio-video response.
For interaction, users can choose voice only, display an avatar beside the chat interface, or switch to a full-screen “face-to-face” mode. Muse’s default avatar is a cartoon character nicknamed “Jolly,” which users can replace with a person, animal, or other character. Hundreds of voice styles are available, including “New Yorker energy” and “pirate style.”
Meta positions this capability as a more natural way to interact with a personal agent: users can assign tasks by voice while walking or when their hands are occupied, and Muse continues working on them in the background. It can also be used for brainstorming, travel planning, and practicing for interviews or speeches. Meta also demonstrated a way to make the Muse avatar dance and share it on Instagram, Facebook, or WhatsApp.