Most AI voice assistants are a disembodied voice and a waveform. Meta is betting that people will bond more with a talking cartoon dog. That’s the pitch behind Muse Voice, which Meta announced on Thursday is now available to everyone in the U.S.
The feature lets users call and video chat with Muse, Meta’s consumer AI agent. It was first shown at the Connect developer conference in September. Two new models drive it: Muse Realtime Avatar and Muse Realtime Voice. Together they power the animated character that talks back to you, nicknamed Jolly by default. You can change it into something more human, an animal, or another character entirely.
One detail matters here. Muse doesn’t actually see you. It hears you and reacts to the conversation, and as it speaks, Meta says the avatar uses “subtle expressions and gestures” that match the tone. So this is a visual layer on a voice product, not a multimodal video understanding feature like the camera modes in ChatGPT or Gemini Live. Don’t confuse the two.
Meta’s logic is plain. If the agent feels like a character, users may feel more connected to it, and that connection should deepen with live conversation. Whether that’s charming or a little manipulative depends on your taste. But it is a real point of difference. OpenAI, Google, and Anthropic all offer voice chat, and none of them gives you a customizable animated face as the default experience. Character-driven AI has been Character.AI’s territory, and Meta now has the distribution to push that idea to a much larger audience.
Alexandr Wang, Meta’s chief AI officer, shared the news in a blog post and on X, along with a demo. In it, he jokes with his Muse, a fluffy white dog, about his habit of posting memes on X getting him in trouble with his boss. The dog suggests adding disclaimers to memes that could be read as Meta endorsing something risqué. It’s a good demo, and a safe one. It shows personality and timing without showing hard task work.
That gap is worth watching. Meta says users can ask Muse to handle tasks for them, but the launch leans on charm, not on proof that the agent completes real work reliably. Agent quality is where competitors have been spending their energy, so the avatar alone won’t carry this.
The more interesting context is the hardware. Meta plans to ship a Tamagotchi-like wearable for Muse before the end of the year, called “Muse Charm.” The avatar will live on its small screen, and users will talk to it and have it reply or act on their behalf. Real-time conversation is supported now on mobile, web, and Mac, and it will later come to Meta’s AI Glasses and the Charm device. Seen that way, this rollout is a software rehearsal for a product category Meta wants to own.
So does it matter? Partly. The avatar is a gimmick if it stays on a phone. But Meta has billions of users, a glasses business, and now a companion device, which is a distribution stack rivals can’t easily copy. OpenAI’s hardware work with Jony Ive is the obvious competitor to watch. If people actually enjoy talking to Jolly, Meta gets a head start on the idea that an AI agent should have a face. If they don’t, it’s a very cute experiment.
