Google DeepMind has introduced a fresh AI model named Genie 3, which can create interactive 3D worlds.

Users simply have to provide a text prompt that describes the environment, and the model will simulate it in real time at 24 frames per second, keeping a steady 720p resolution for several minutes.

Last year, the team launched their initial “world models”, Genie 1 and Genie 3.

On top of that, cutting edge AI video generation models such as Veo 2 and Veo 3 also show a grasp of the physical world.

A blog that accompanied the release mentioned that these world models, capable of comprehending environments and recreating them, assist agents in predicting how changes in the environment occur and how their actions can influence it.

The team mentioned that while Genie 2 had an interactive window lasting around 10 to 20 seconds, Genie 3 provides interaction for a “few minutes”.

Additionally, the AI model is better at maintaining visual consistency, so if a user leaves a location and comes back later, it will look the same.

That said, Genie 3 isn’t open for public preview just yet and will be introduced to a limited number of creators for testing.

Subscribe My Channel





Discover more from Connect2ConnectOnline

Subscribe now to keep reading and get access to the full archive.

Continue reading