Google Genie 4: AI model for interactive world simulation in real time
In brief
- Genie 4 generates interactive environments in real time with on-the-fly modifications.
- The model produces sound and maintains a coherent scene for about two minutes.
- Potential applications: robotics, gaming, and audiovisual previsualization.
Google DeepMind unveils Genie 4, an AI model capable of generating interactive environments in real time, with on-the-fly modifications and integrated sound.
The model allows users to adjust a scene while it runs using natural language instructions, such as changing the weather or adding an object, while maintaining its coherence. Genie 4 also generates ambient sounds and dialogues, leveraging capabilities from the Gemini family to improve complex event handling. According to TechCrunch, it can sustain an interactive scene for about two minutes—double the duration of Genie 3—but this remains limited for advanced use cases.
Potential applications span robotics, where simulations could test agent behavior before real-world deployment, as well as gaming and audiovisual previsualization. Google DeepMind is currently testing the model with a small group and plans wider access via Project Genie and Google AI Ultra, with no general deployment date announced. Current limitations include object stability, text readability in scenes, and long-term physical interaction coherence.
What it changes for you
- Video GamesGame developers could accelerate prototype creation using interactive simulations before final development.
- Media & EntertainmentFilm crews could quickly test shots, lighting, and movements through interactive previsualizations.
- Industry & ManufacturingSimulations could help validate the behavior of robotic agents before real-world deployment.
This summary gives you the essentials. The full article, its figures and quotes are on Le Monde Informatique : IA.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error