Google DeepMind has released Genie 4 as a research preview, presenting a world model that generates more photorealistic and interactive scenes and produces real-time audio that matches the simulated environment. The model accepts freeform prompts during a running simulation, enabling on-the-fly edits such as changing weather, camera angles, or adding people and objects.
Capabilities and integrations
Genie 4 doubles the sustained consistency of its predecessor by maintaining a scene for up to two minutes and introduces improved physics and more human-like characters. It also generates synchronized audio — music, ambient sounds, and interactive dialogue — intended to align with the visual simulation.
The model works more closely with Gemini, which acts as a director to plan steps and check frames while Genie creates visuals. In a research experiment tied to Gemini Robotics, a Gemini 4 Argon model decomposed tasks into steps that Genie 4 simulated in real time, illustrating a method for testing robot scenarios in simulation before real-world execution.
Use cases, industry context, and limits
Google and others pitching world models point to use cases such as gaming, interactive learning, and robotics. A speculative example offered is entering a historical painting and conversing with its subjects. Diego Rivas, a Google DeepMind engineer, said the real-time aspect could be valuable during ideation and storyboarding, and that filmmakers have begun using Veo for high-definition clips with synchronized audio.
Tejas Kulkarni, research director at Google, described a future of “hybrid world models, where you have a combination of Gemini writing code and text, and Genie ingesting that in real time and simulating reality at different levels of abstraction.” Rivas added that turning prototypes into finished creative assets via Gemini Omni remains some way off.
Teams acknowledged current gaps: details and objects can drift in longer runs, audio accuracy can be inconsistent, and in-world text may be illegible. The release follows other recent entries such as Runway’s GWM Worlds 2 and comes amid consolidation in the sector — examples cited include the acquisition of World Labs by AMD, Amazon’s talks to buy Decart, and Rivian’s purchase of Third Dimension.
Genie 4 is available now as a research preview to trusted testers and will roll out in the coming weeks to Google AI Ultra subscribers in Project Genie and to early access partners of the Gemini API and Google AI Studio.
Original source: TechCrunch AI