Google Launches Project Genie in US, an Interactive World Generation Prototype Powered by World Model Genie 3

This prototype is a web application powered by the world model "Genie 3," the image generation model "Nano Banana Pro," and the large language model "Gemini." Users can construct dynamically expanding environments using text prompts or images, exploring them while interacting in real time.

Unlike traditional static 3D snapshots, Genie 3 generates subsequent paths in real time based on the user's movement and actions. Google stated that because physical laws and interactions between objects are also simulated, the system can reproduce a wide variety of real-world scenarios, ranging from robotics to animation production.

To provide more precise control, a "World Sketching" feature has been integrated. This allows users to modify images and fine-tune aspects such as the perspective (first-person or third-person) before entering the world. Additionally, users can remix existing worlds or draw inspiration from a gallery.

Google explained that some features of Genie 3 announced in August, such as promptable events that change the world during exploration, are not yet included in this prototype.