Meta says new avatar model lets users video chat with Muse agent
Alexandr Wang said Muse Realtime Avatar animates the agent while it speaks and answers in under a second, with every output watermarked.
Meta said users will soon be able to video chat with their Muse agent, driven by a new model it calls Muse Realtime Avatar. Meta's Alexandr Wang said the model animates the agent while it talks, and that voice and video stay in sync.
TechCrunch, which covered Meta's Connect keynote in Menlo Park on Wednesday, said users can give the agent a distinct digital avatar and then video call it. TechCrunch said Meta gave no date beyond saying the feature is coming. Wang said the agent answers in under a second, for as long as a user keeps talking.
Wang said Meta put the model head to head against Runway Characters and HeyGen LiveAvatar, using each product's own live-call experience with the same character on both sides, and that people preferred Muse. That test is Meta's own, run on a comparison it chose, and no one outside the company has repeated it.
Everything the model makes is watermarked as AI without adding latency, Wang said. He called it the foundation for real-time, embodied AI across Meta's products. Meta has not published a model card or a paper, so the latency and preference figures could not be checked outside its own demonstrations.
Mark Zuckerberg said Meta packed the real-time voice and avatar stack into Muse Charm, a keychain device with a small screen and a fingerprint sensor. He said it should ship in December, gave no price, and said the team still had to finalise the components, according to TechCrunch.