Google DeepMind released its newest world model, Genie 4, on Tuesday as a research preview.
The latest model generates the most realistic and interactive worlds to date, allowing users to input prompts during a simulation to change the scene.
The model can also generate real-time audio that matches the simulated world, including music, ambient noise, and interactive dialogue.
The launch of Genie 4 comes just months after AI video startup Runway released its second world model, GWM Worlds 2, a playable model that responds dynamically to user input and includes audio generation.
Meanwhile, the world model industry is beginning to consolidate.
Fei-Fei Li's World Labs is being acquired by AMD, Amazon is in talks to acquire Decart, and Rivian has just acquired Third Dimension to bring simulation capabilities in-house for autonomous driving testing.
Genie 4 is Google's latest move to stay ahead in a crowded field.
Google released the previous generation, Genie 3, as a research preview about a year ago.
Beyond being far more realistic than its predecessor and featuring audio capabilities, Genie 4 also allows free-form editing, such as changing the weather or camera angle in real time, or adding new characters or objects.
It can maintain a consistent scene for up to two minutes, double the previous generation's limit, and offers better physics and more human-like characters.
Google's AI model Gemini is also more tightly integrated with Genie 4, acting as a "director" that plans steps and checks the visuals, while Genie handles generating the visual content.
In a research experiment related to Gemini Robotics, the Gemini 4 Argon model broke tasks into steps that Genie 4 simulated in real time.
The experiment demonstrated a possible approach: testing robotic scenarios in simulation before they happen in the real world.
"In the future we will increasingly see this kind of hybrid world model, where Gemini writes code and text, and Genie receives and simulates reality at different levels of abstraction in real time," said Tejas Kulkarni, a director of research at Google, during a briefing.
As with other world models, the touted commercial use cases include gaming, interactive learning, and robotics.
One speculative example of interactive learning is stepping into a painting of a historical scene and talking with the figures in it about the events depicted.
But "trusted testers," the first people to gain access to Google's cutting-edge products, are pointing to a possible new use: film prototyping.
"Real-time is most valuable where it's very useful for maintaining a flow state in creative fields," said Diego Rivas, an engineer at Google DeepMind, during a news briefing.
"For example, in the ideation phase, the storyboard phase, any time you need to quickly generate ideas."
Google also offers Veo, its AI video generation model, which can create high-definition clips with built-in synchronized audio from text, image, or video prompts.
Rivas added that users may eventually rely on Gemini Omni to turn these prototypes into creative assets.
However, he said those capabilities are still some distance away.
Throughout the briefing, Rivas emphasized that Genie 4 is still in its early stages and that it is not yet clear what the end-to-end product experience will look like.
The team also acknowledged that Genie 4 still has gaps: the longer a simulation runs, the more details and objects may drift, and they have not fully figured out how to ensure audio accuracy and clearly readable text within the world.
Genie 4 is currently available to trusted testers as a research preview.
In the coming weeks, the model will be made available to Google AI Ultra subscribers for Project Genie and to early access partners on the Gemini API and Google AI Studio.
Comments