Google DeepMind’s Genie 4 generates sound alongside interactive images and lets users change a scene while exploring it. A restricted research preview is available to trusted testers. Wider consumer and developer access is planned, while the company acknowledges problems with persistent objects, complex physical interactions and the consistency of longer sessions.
Google DeepMind introduced Genie 4, an artificial intelligence model that creates interactive simulated worlds, on October 27. Users can explore the generated environment using keyboard controls and move the camera while sounds accompany the images. Ambient noise, music and character dialogue are generated within the same experience. Text instructions can change the weather or introduce a person or object as the scene unfolds.[1], [2]
Genie 4 incorporates several models from Google’s Gemini family. Google says Gemini Omni provides a foundation for appearance and behaviour, while Gemini 3.8 Flash observes generated frames, breaks goals into steps and checks their progress. The plan can adapt when a user intervenes. Film previsualisation and interactive game prototypes are among the uses being explored with testers from creative industries, education, journalism and engineering.[1]
Access remains a research preview for trusted testers. Google plans subsequent access through Project Genie for Google AI Ultra subscribers and early developer partnerships. The company acknowledges that objects can change appearance or position during longer sessions and complicated edits. Complex physics, accurate sound and readable text remain research problems. Both video and audio outputs carry SynthID, Google’s watermark for identifying generated content.[1]
