Imagine playing a game where the world you're exploring is being generated as you move
That's what Google DeepMind is trying to build with Genie 4
You can walk through a photorealistic environment, change the scene with text prompts, and interact with characters that respond with speech
The model also generates sound and simulates physical interactions, from collisions to movement
Genie generates the world while Gemini helps keep events coherent, plans multi-step actions and adapts as you interact with the scene
That makes it more than a video generator
A video is something you watch
A world model generates an environment that responds to what you do inside it
DeepMind sees applications in game prototyping, creative work and training AI agents or robots in simulated environments before they operate in the real world
But Genie 4 is still an experimental research preview
Objects and visual details can drift over time, and complex physical interactions aren't always consistent
The bigger idea is that AI could eventually generate environments you don't just watch, but explore, change and interact with in real time
JUST IN: Google DeepMind unveils Genie 4, a new AI model that generates photorealistic, interactive 3D worlds in real time, complete with physics, sound, & human dialogue.