Nvidia wants to fill the virtual and physical worlds with AI avatars

Nvidia announced Omniverse Avatar, a platform that allows virtual agents to be created. Nvidia claims that the platform can power a variety of virtual agents by combining a variety of technologies, including speech recognition, facial tracking and 3D avatar animation.
Jensen Huang, Nvidia CEO, showed a few demos of Omniverse Avatar technology during a presentation at the annual GTC conference. One of the demos features a cute animated character that talks to a couple using a digital kiosk. He answers questions such as which items are vegetarian. To maintain eye contact with customers and respond to facial expressions, the character uses facial-tracking tech. Huang said that the technology will be beneficial for smart retail, drive throughs and customer service.

Another demo featured an animated toy Huang answering questions on topics such as climate change and protein production. In a third demonstration, someone used an animated avatar of themselves to act as a stand in during a conference call. Although the caller was casually dressed in a busy cafe environment, their virtual avatar was smartly dressed and spoke clearly without any background noise. This is an example of Nvidia's Project Maxine project, which aims at improving common videoconferencing problems (such as low quality streams or maintaining eye contact) through machine learning.

You can see Huang's toy version in the video below. It starts at 28 minutes. You can also skip ahead to 1 hour and 22 minutes for the demo kiosk.

Nvidia's "omniverse vision" includes the announcement of the Omniverse avatar -- a bold bit of branding for a complex collection of technologies. The "omniverse," like the "metaverse," is essentially about sharing virtual worlds that enable remote collaboration. Nvidia's vision is different than that of Meta, the Facebook-owner. It is more concerned about creating virtual environments and interacting with physical people.

Nvidia's presentations looked great, as usual with these presentations. However, it isn't clear how practical this technology will actually be in the real world. It's unclear whether customers will prefer interactive kiosk experiences to selecting items from a menu. Huang pointed out that the avatar's response time is two seconds -- slower than a human and likely to cause frustrations for customers who are in hurry. We have yet to see Project Maxine technology in action in real life, even though it looks impressive.