Google DeepMind announced Genie 4 on Oct. 27, 2026, opening a restricted research preview of an AI world model that generates interactive environments and sound. The release gives creative and engineering teams a new simulation tool to evaluate, but not a generally available production service.
In its announcement, DeepMind says the model generates music, ambient sounds and dialogue alongside visual scenes. Users can steer a running simulation with text, including changing weather or introducing a person or object.
Access remains restricted
DeepMind says trusted testers have begun receiving the research preview. Access for Google AI Ultra subscribers through Project Genie and early-access partners using the Gemini API and Google AI Studio is coming later. The company offers a waitlist rather than unrestricted access.
TechCrunch also reported the research-preview release Tuesday.
Consistency is still a limitation
DeepMind acknowledges that visual details and object permanence can drift, particularly during longer sessions or complex edits. Physical interactions, audio accuracy, readable text and spatial consistency remain research challenges. The company says generated video and audio will include its SynthID watermark.
For prospective business users, the distinction matters: an interactive demonstration can support experimentation without establishing that its simulated events accurately predict real-world outcomes.
A team assessing the preview would therefore need to separate creative exploration from engineering validation. A convincing scene is not, by itself, a measurement of how equipment, people or processes will behave.