Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering

THE NEWS

What happened

Google DeepMind has released Genie 4, a world model that generates photorealistic, interactive simulations with synchronized real-time audio. The model supports free-form edits like changing weather or camera angles mid-simulation and can maintain consistent scenes for up to two minutes. Integrated with Gemini, it aims to accelerate robotics testing, film prototyping, and creative ideation, though it currently faces challenges with object drift and text legibility.

CONTEXT

Why it matters

Google DeepMind has unveiled Genie 4, its latest world model capable of creating photorealistic, interactive simulations with synchronized real-time audio. The system allows users to edit scenes in real time—changing weather or camera angles—and maintains consistency for up to two minutes. Integrated with Gemini, it aims to revolutionize robotics testing and creative workflows like film prototyping. Currently available as a research preview.

AT A GLANCE

Key facts

• Genie 4 generates photorealistic interactive worlds with synchronized real-time audio including music and dialogue.
• The model allows free-form edits such as changing weather or camera angles in real time during simulation.
• Scene consistency is maintained for up to two minutes, double the limit of its predecessor Genie 3.
• Genie 4 is integrated with Gemini, which acts as a director to plan steps and check frames for robotics testing.
• Trusted testers highlight potential uses in film prototyping and rapid creative ideation during storyboarding.
• Current limitations include object details drifting over time and issues with accurate audio and legible in-world text.
• The model is currently available as a research preview for trusted testers and will roll out to Google AI Ultra subscribers.