Google has announced that it has already opened access to Project Genius so that Google AI Ultra subscribers can try this experimental research prototype, which allows you to create and explore three-dimensional worlds in real time.
Genie 3 is a dynamic world generative artificial intelligence that has already proven to outperform its predecessor, as well as GameNGen and Veo. Unlike other generative video AIs, like Sora or V1, he not only creates spaces to be seen, but also to be explored.
In this way, it allows users to create video games and three-dimensional scenarios in real time, that they can visit by moving with the keyboard cursors, as in a video game, as well as interact with the objects in that world.
A model that simulates the dynamics of an environment
Google has stuck out its chest explaining what, so far, Genie 3 is capable of, insisting that «unlike explorable experiences in static 3D snapshots», its model «generates the path forward in real time as you move and interact with the world.»
Based on his research, he has taken a step forward with the Project Genie prototype, a web application prototype powered by Genie 3, Nano Banana Pro and Gemini. This allows users to experience first-hand the immersive experiences of their model.
In turn, this experience focuses on three main capabilities. the first, the one known as Sketch of the worldwhich has been integrated with Nano Banana Pro and allows you to preview the appearance of the world you want Create and modify the image to fit it before you start exploring it.
Another essential aspect of Genie 3 is what is known as World explorationwhich causes the path that the user takes to be generated in real time as they move through the different scenarios.
Finally, Google allows you to carry out a kind of remixing of existing worlds, to give them new interpretations. Likewise, you can explore selected scenarios and worlds from the model gallery or click on the randomization icon.
In this way, users They can be inspired by other worlds already created to develop your own or start building them from them before downloading them.
For now, only 60 second worlds
Google has recalled that Project Genie is an experimental research prototype at Google Labs that, at the moment, is in the testing phase and may still contain some errors.
Therefore, he has pointed out that generated worlds may not appear completely real nor do they always adhere to the instructions or images provided by users.
Likewise, it is likely that the characters may be less controllable or experience higher latency, so that they do not respond when the user wants to give them a certain movement.
Finally, Google has remembered that videos of interactive and three-dimensional worlds that allows this artificial intelligence to be built are limited to 60 seconds; without specifying whether it will extend this duration later.