Google DeepMind is making AI-generated virtual worlds considerably more interactive with Genie 4, its latest world-generation model. The new model brings real-time native audio and the ability to modify environments while exploring them, allowing users to change what happens inside a generated world without starting over.
Announced on November 18, 2026, Genie 4 is arriving as part of an expanded Project Genie experience, Google’s experimental platform for creating and exploring interactive digital environments. The update also introduces a conversational creation agent, a community gallery, and several improvements to the existing Genie 3 model.
The biggest change is how users interact with worlds after they have been generated. Previously, creating an environment was largely driven by an initial text or image prompt. With Genie 4, users can now issue additional natural-language instructions during an active session to change the environment in real time.
Genie 4 brings real-time audio and world editing
Genie 4 allows users to reshape an environment while moving through it. For example, someone exploring a generated landscape can change the weather, introduce a new character, or redirect the action simply by typing another prompt.
The model also introduces real-time native audio, pairing generated visuals with sounds that respond to what’s happening in the environment. This includes ambient soundscapes and character dialogue, making generated worlds feel more complete than visually interactive environments alone.
Google describes Genie 4 as its most realistic and interactive world model yet, with improvements to visual fidelity and physics alongside its new audio capabilities.
The ability to edit an environment mid-session is particularly interesting because it moves AI world generation beyond producing a predetermined scene. Users can influence how their surroundings evolve, effectively turning exploration into an ongoing creative process.
Project Genie gets a conversational creation agent
Google is also introducing a new Project Genie agent that makes building interactive experiences less dependent on carefully constructing individual prompts.
Instead of describing an entire environment through a single instruction, users can explain their ideas through a natural-language conversation. The agent coordinates Google’s multimodal models to generate the initial visuals, establish the world’s visual identity, and create background music suited to the intended atmosphere.
Gemini models can also participate during exploration. They can observe what’s happening on screen, suggest relevant next steps through contextual suggestion chips, and automatically trigger events within the generated environment.
This creates a more collaborative experience in which users can build, explore, and modify virtual worlds through conversation rather than relying entirely on traditional creative tools.
Google is highlighting several examples of what the technology can produce. These include Villa Giallo, an interactive mystery set inside a gothic manor, and The Life of Ants, an experience created with filmmaker Anand Gandhi that lets users explore a colony from an ant’s perspective.
Genie 3 gets longer sessions and 360-degree environments
Genie 4 isn’t the only model receiving attention in this update. Google is retaining Genie 3 in Project Genie’s model picker and adding capabilities that make it useful for different types of experiences.
One notable improvement is extended generation sessions. Genie 3 can now support up to three minutes of exploration per generation, compared with the previous 60-second limit. Users can also connect multiple generated worlds together, creating longer journeys across different environments.
Another addition is 360-degree panorama image grounding. Users can provide panoramic images as starting points for generated environments, helping the model maintain visual consistency with the reference surroundings.
Genie 3 also gains real-time in-world events, allowing users to modify environments through natural-language instructions while exploring. This makes interactive changes available in both models, although Genie 4 adds higher-fidelity visuals, physics, and native audio.
A community gallery makes worlds easier to share
Project Genie is also becoming more social with the introduction of a community gallery. Users can browse curated environments, explore creations from other people, and remix existing experiences.
The gallery includes projects such as Living Memory, which turns personal photographs into explorable environments, and Dino Field Guide, which uses 360-degree panoramas to create prehistoric habitats. Another example, Rock Paper Scissors, places themed objects on a desk with Gemini acting as a real-time referee.
Users can also publish their own creations for others to explore and modify. Beyond changing the appearance of an existing environment, remixing can involve adjusting its underlying structure, soundtrack, and rules.
This approach gives Project Genie a community-driven dimension, where generated worlds can become starting points for new creative experiments rather than remaining isolated personal projects.
Project Genie availability and safety
The updated Project Genie experience, including Genie 4, is rolling out to eligible Google AI Ultra subscribers. Google AI Pro members can join a waitlist through the Project Genie website.
Google says all video and audio generated through Project Genie includes SynthID watermarking, while model inputs and outputs undergo safety screening.
However, Genie 3 and Genie 4 remain frontier research technologies rather than finished world-building platforms. Google acknowledges that challenges remain around camera latency, complex physical interactions, real-time audio accuracy, and maintaining scene consistency over extended periods.
The new capabilities nevertheless point toward a more flexible approach to generative AI. Rather than simply creating images or short video clips from instructions, world models are beginning to produce environments that users can enter, interact with, and continuously reshape.
With Genie 4, Google is bringing that idea closer to a conversational creative experience, where the world doesn’t have to remain the way it was originally generated.
Discover more from GadgetBond
Subscribe to get the latest posts sent to your email.
