Google announces Genie 4, limited to a trusted-tester research preview
The world model generates interactive scenes and sound; wider Ultra-subscriber and early-access API availability is still planned.
Google DeepMind announced Genie 4 on October 27, reporting a research preview for trusted testers; the exact activation time is unknown. Its demonstrations show real-time scene editing and native audio, with acknowledged limitations in objects, text and longer-term consistency.
Project Genie access for Google AI Ultra subscribers and Gemini API/AI Studio early-access partners is described as coming soon, not current access for every paying user. Hong Kong, Macau and Taiwan eligibility, Traditional Chinese support, fees and usage limits are unconfirmed. General web, mobile-app and public API availability is not established. Not tested.
Project Genie access for Google AI Ultra subscribers and Gemini API/AI Studio early-access partners is described as coming soon, not current access for every paying user. Hong Kong, Macau and Taiwan eligibility, Traditional Chinese support, fees and usage limits are unconfirmed. General web, mobile-app and public API availability is not established. Not tested.
Usage conditions
- Hong Kong
- Hong Kong eligibility, Traditional Chinese, pricing and rollout date are unconfirmed; an Ultra subscription does not establish access.
- Macau
- Macau eligibility, Traditional Chinese, pricing and rollout date are unconfirmed; an Ultra subscription does not establish access.
- Taiwan
- Taiwan eligibility, Traditional Chinese, pricing and rollout date are unconfirmed; an Ultra subscription does not establish access.
Official source: Genie 4: A new generation of world simulation