Google announced today that its Genie 4 world model can generate interactive experiences lasting up to two minutes, an increase from the previous one-minute limit. The system relies on Gemini 3.8 Flash to power simulations and introduces four specific updates. Users can now request more realistic environments, modify scenes instantly using natural language prompts, and include audio generation within the output. The company states the tool aids pre-visualisation for filmmakers, prototypes game concepts, and creates training scenarios for robots.
While the extension of duration and audio capabilities represent tangible improvements for simulation tasks, the technology remains a research prototype rather than a public product. No changes are currently confirmed for customer-facing applications, API behaviour, or data privacy policies. The model demonstrates progress in world simulation but does not yet offer a functional interface for external developers.
- Video clips run for a maximum of two minutes
- Users can edit scenes with text prompts
- Audio is generated alongside visual content
