Google launches Project Genie, its AI prototype to generate interactive worlds

Google has announced the launch of Project Genius a prototype that combines technology Genie 3, Nano Banana y Gemini to generate interactive worlds with artificial intelligence.

In August, Google DeepMind introduced the Genie 3 model capable of generating more realistic dynamic environments with real-time interaction for several minutes with a resolution of 720p, surpassing its previous version, limited to static 3D images.

After being restricted to developers, Google has chosen to offer users the opportunity to try Genie 3 through Project Genius a world generator prototype that combines this technology with the generation of images of Nano Banana and the assistant Gemini.

In this way, users can “live first-hand the immersive experiences of these world generators,” Google explained on its website. DeepMind Research Director Shlomi Frutcher explained in an interview with TechCrunch that it is “exciting to be in a place where more people can access and give their opinion.”

Project Genie works with three features: a world sketch, exploration, and world merging. First of all, Nano Banana Pro allows preview the world in a sketch before generating it through the user’s instructions (creation of the environment, characters, and the method to explore it, among others).

Once the world is created, the player can explore it freely thanks to the generation of environments in real time according to the user’s actions. Lastly, the player can also reedit and combine worlds generated previously to customize them again.

For now, Google has limited the use of Project Genie to Google AI Ultra subscribers in the United Stateswith a maximum duration of 60 seconds generation and navigation. “The reason we limited it to 60 seconds is because we wanted it to reach more users,” Fruchter clarified. “When you use it, there is a chip somewhere that is just yours and is dedicated to your session.”

In addition, Google has warned that being a prototype, the generated worlds may not appear “totally real or always adhere to the instructions”, the control of the characters may present errors and there may be greater latency during the experience.