Trusted testers to explore how Google’s next Genie AI model generates realistic experiences with audio
What Genie 4 can do
Google says Genie 4 has four main updates: more realistic worlds, real-time changes prompted in natural language, generated audio and simulations powered by Gemini 3.8 Flash. The model can create experiences of about two minutes, compared with one minute for the previous Genie 3 model. The company says the worlds are interactive.
Potential uses and limits
Google describes possible uses including pre-visualization for filmmakers, prototyping game ideas and creating training scenarios for robots. The company cautions that details and object permanence can shift as simulations run longer.
Testing and availability
Genie 4 is launching first as a research preview for trusted testers. It will be coming soon to Google AI Ultra subscribers, and Google has opened a waitlist for people seeking access through Project Genie.
The reporter had not tried Genie 4 and could not compare it directly with Project Genie, which uses Genie 3. Google shared examples of what the new model can create.
Key points
- Genie 4 can create interactive experiences lasting about two minutes.
- Google says the model can generate audio and accept real-time natural-language changes.
- Potential uses include filmmaking pre-visualization, game prototyping and robot training.
- Longer simulations may see details and object permanence shift.