Skip to main content

Google introduces Genie 4 for interactive world simulation

18 NOVEMBER 2026·2 MIN READ·6 SOURCES

Google DeepMind introduced Genie 4, a world-simulation model, in a post dated October 27, 2026. The company says it improves realism and real-time interaction, and describes uses ranging from creative tools to robot and agent training environments.

Key takeaways · 3

  • 01

    Users can make real-time scene changes with natural-language prompts, including changing weather, moving the camera or adding characters.

  • 02

    Google says Genie 4 generates native audio, including ambient sounds, music and interactive dialogue.

  • 03

    Media reports describe experiences of about two minutes, but also report that worlds can lose context and that objects or details may drift.

A world model users can steer

Google DeepMind says Genie 4 creates simulations that respond after an event is set in motion, including how people move, react and speak.[2] Google says Gemini Omni grounds the simulations in real-world appearance and behavior, while other Gemini models help maintain coherence and continuity.[1] The company describes improved photorealism and fluid movement, and says the model is intended for robot and agent training environments as well as creative tools.[2][1] These are stated capabilities and intended uses, rather than evidence of performance in a particular training or production workflow.[1][2]

Real-time controls and generated audio

Users can navigate worlds with standard keyboard controls and steer scenes with text prompts, according to Google DeepMind.[2] Google says prompts can change weather, move the camera or add characters, and that the model can contextualize changes to fit the scene and continue steering between user inputs.[1][2] Genie 4 also generates audio alongside its worlds, including ambient sounds, music and interactive dialogue; Google DeepMind says the sounds in its demonstration came from the model itself.[1][2] Google DeepMind’s November 18 post also highlighted native audio, real-time editing and a new agent.[6]

Duration and reported limitations

The Verge and TechCrunch report that Genie 4 can maintain a scene for up to about two minutes, twice Genie 3’s reported one-minute limit.[3][4] Ars Technica reports that worlds may lose track of their original context after a couple of minutes.[5] TechCrunch reports that simulation details and objects can drift, and that audio accuracy and in-world text legibility remain unresolved.[4] These reports make duration and consistency practical factors to assess before teams rely on a generated world for a longer or detail-sensitive task.[3][5][4]

Preview access and possible uses

Google DeepMind says Genie 4 has started rolling out as a research preview to trusted testers, while Google says it is coming to Google AI Ultra subscribers through Project Genie and to early-access Gemini API and AI Studio partners.[2][1] Google DeepMind invited users to join the Project Genie waitlist and later said Genie 4 was available there.[2][6] Ars Technica reports that Google AI Ultra plans start at $100 per month; that is the reported plan price, not a separately stated Genie 4 price.[5] Google has pitched the model for filmmaker pre-visualization, game prototyping and robot training.[3]

For teams exploring simulation, prototyping or visual ideation, Genie 4 offers a way to test how a scene changes in response to prompts rather than treating it as a fixed image. The reported short duration and consistency issues make it worth testing against the length and precision requirements of a real workflow before building plans around it.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 18 November 2026

    Google introduces Genie 4 for interactive world simulation

Sources

AI fluency, one session a day, built for your work.