AI News
AI News AgentProductGoogle3 min read

Google opens Genie 4 for creating interactive worlds

Google is adding `Genie 4` to Project Genie, an experience that lets you create and modify interactive worlds with text instructions and real-time generated audio. Access begins for eligible Google AI Ultra subscribers, while AI Pro users can join a waitlist.

Google now lets you create and explore interactive worlds with Genie 4, although access is currently limited to certain Google AI Ultra subscribers. The update arrives as part of Project Genie, an experience that combines world-generation models with Gemini models so you can build, modify and share environments from your browser.

The idea is simple: you write what you want to see and enter that world to explore it. You can then keep giving instructions as you move through it. For example, you can change the weather, add a character or alter what happens in the scene without starting over.

Genie 4 changes the world as you explore it

Google presents Genie 4 as its most realistic and interactive world model to date. Its main addition to Project Genie is the ability to edit the environment during a session using instructions written in natural language.

It also includes native real-time audio. The sound adapts to what is happening in the scene, from ambient sounds and effects to character dialogue. If you walk through a forest, the environment can respond with sounds that fit the situation. If a character appears, they can speak in their own voice.

The goal is not to create a finished film, but a space that responds to your decisions. That could help you prototype a story, design a visual experience or simply explore an imagined setting without programming skills.

Genie 3 gets longer sessions

Genie 3 will remain available through the Project Genie model selector. According to Google, it may be more useful when you want longer sessions or want to create spaces based on panoramas.

Its main updates are:

  • Sessions of up to three minutes per generation, compared with the previous 60-second limit.
  • The ability to chain multiple worlds together to create longer journeys.
  • Creating environments from 360-degree panoramic images.
  • Events that change in real time as you explore, triggered by text instructions.

An agent helps build the experience

Project Genie also includes an agent that turns a conversation into a first interactive world. You can describe an idea, and the agent coordinates several of Google's multimodal models to generate the initial appearance, define the style and create a soundtrack that fits the atmosphere.

Gemini models can participate during exploration. They observe what is happening on screen and offer suggestions about what to do next, or automatically trigger events within the world.

The platform also includes a gallery of worlds created by Google, collaborators and the community. Examples include spaces built from personal photos, prehistoric habitats based on 360-degree panoramas and interactive games in which Gemini acts as the referee.

You can also modify and reuse existing worlds. You can change their structure, music or rules and publish the result so other people can explore or transform it.

Limited access and technology that is still experimental

The new Project Genie experience is available starting today to eligible Google AI Ultra subscribers. Google AI Pro members can join a waitlist.

Google acknowledges that Genie 3 and Genie 4 are still in the research phase. The company is continuing to work on issues such as camera latency, complex physical interactions, real-time audio accuracy and the ability to keep a scene stable for long periods.

All generated audio and video includes SynthID watermarks, and inputs and outputs go through safety checks. For now, the important point is not that these worlds will replace video games or film, but that they bring interactive environment creation within reach of anyone who can describe an idea. The next thing to watch will be whether Google can maintain that editing freedom without losing coherence as sessions become longer and more complex.