Every video I make follows the same honest routine: I generate a short clip, the audience watches it, and that is the end of the road. Nobody can step inside it. Nobody can ask the character a question and hear an answer. That is why Google DeepMind's announcement of Genie 4 this week stopped me mid-scroll. This is not another video generator. It is a world model, a system that builds a place you can move through in real time, and for the first time, that place has a voice of its own.
Let me explain the difference, because it matters. A normal AI video model renders a clip and finishes. A world model simulates an environment as you explore it, predicting what you will see, hear and bump into next. DeepMind says Genie 4 is natively multimodal, which means the music, the ambient noise and even spoken dialogue are generated together with the world itself, not pasted on afterwards. You set an event in motion, and the model simulates how the world responds, how it sounds, and how the people in it move, react and speak.
The part that excites me most is the steering. While a simulation is running, you can type a prompt and change the scene on the fly: shift the weather, move the camera, drop in a new person or object. DeepMind says a scene can now stay consistent for up to two minutes, double the limit of Genie 3, with better physics and more humanlike characters. Under the hood, the Gemini family does the directing. Gemini Omni anchors the simulations in real-world appearance and behaviour, while other Gemini models supply the reasoning that keeps a scene coherent, more like a director planning steps and checking frames while the world renders around you.
DeepMind is aiming this at serious work, not just play. The company talks about safe, endlessly variable simulations for training robots and AI agents, and about unlocking entirely new interactive experiences. The timing tells its own story. Runway released its second world model, GWM Worlds 2, a couple of months ago. Fei-Fei Li's World Labs is being acquired by AMD, and Amazon is reported to be in talks to buy Decart. When the giants start buying the mapmakers, you know the territory is valuable.
Now the honest limits, because this blog does not sell dreams. Genie 4 is a research preview for trusted testers only. The wider rollout comes later, starting with Google AI Ultra subscribers in Project Genie and early access partners on the Gemini API and Google AI Studio, and there is a waitlist if you want in. Even DeepMind admits the details in a generated world can still shift and drift over time. I cannot open it tonight and build tomorrow's cartoon inside it, and neither can you.
But I know which way the river is flowing. My craft today is stitching eight-second clips together and hoping the story survives. A world that holds together for two minutes, with characters who converse while you walk through it, is a glimpse of stories you do not just watch but inhabit. When this opens up, the winners will be the storytellers who were ready. I intend to be one of them.
Haroon Ch
Ai Game Changer Pro
Top comments (0)