Google released Genie 4 on November 18, 2026, its latest AI world model that allows Google AI Ultra subscribers to generate and interact with virtual worlds in real time. This update introduces mid-session environment editing, native adaptive audio, and a new community gallery for sharing creations.
Project Genie is an experimental research prototype that lets users create and explore interactive digital environments. The update aims to give users deeper control, easier building tools, and new ways to share their AI-generated worlds. This rollout enhances the interactive nature and realism of AI-powered world creation.
What is Google's New Genie 4 Model?
Genie 4 is Google's most advanced world model, offering enhanced realism and interactivity for generating digital environments. Users can now modify worlds in real time using text prompts, changing weather or adding characters as they navigate. It also integrates native audio that adapts dynamically to on-screen events.The model lets users type prompts mid-session to redirect action within the created world. This means environments can shift instantly, from sunny skies to stormy weather, or new characters can appear on command. Genie 4 also pairs these visuals with real-time native audio, which includes ambient soundscapes and character dialogue that changes with the scene.
Google states that Genie 4 uses Gemini 3.8 Flash for its simulations, supporting more complex and realistic interactions. This integration helps the model deliver higher-fidelity visuals and physics within the generated worlds.
How Does Genie 4 Improve World Creation?
Genie 4 significantly improves world creation by offering more realistic and interactive experiences, extending session lengths, and introducing new creative agents. It builds on previous models, allowing for greater creative freedom and social interaction through a new community gallery for sharing custom worlds.While Genie 4 provides high-fidelity experiences, Google also updated Genie 3. Genie 3 now supports extended sessions of up to three minutes per generation, up from 60 seconds. It also features 360-degree panorama image grounding, allowing users to seed new worlds from panoramic images. Real-time in-world events can dynamically change the environment with natural-language prompts.
| Feature | Genie 3 (Updated) | Genie 4 (New) |
|---|---|---|
| Max Session Length | Up to 3 minutes | Up to 2 minutes |
| Real-time Editing | Yes, with prompts | Yes, with prompts |
| Native Audio | No | Yes, adaptive audio |
| Panorama Grounding | Yes | No |
| Visual Fidelity | High | Highest, most realistic |
A new Project Genie agent simplifies the creation process. Users can describe their vision in natural language, and the agent orchestrates various Google multimodal models to craft initial visuals and custom background scores. This conversational approach makes building complex environments more accessible. Creations can be published to a community gallery, allowing others to explore and remix.



