Google Opens Genie 4 World Model With Real-Time Editing and Native Audio

Original title: 史诗级进化!谷歌 Genie 4 世界模型开放体验:支持边玩边改天气、召唤角色,所见皆可互动
78AI Score

Google has opened a new Project Genie experience to qualifying Google AI Ultra subscribers, granting first access to the Genie 4 world model. Genie 4 lets users edit an existing world mid-session via natural-language prompts (switching weather, adding characters) and debuts native real-time audio that adapts to the scene; Genie 3 also gains up to 3-minute sessions, 360-degree panorama grounding and world chaining. All outputs carry SynthID watermarks, and the models remain frontier research.

Full Report

IT之家 November 19 news, earlier this year, Google launched the experimental research prototype Project Genie, which allows users to generate and explore interactive virtual worlds. Early test users reported that they wanted stronger control over what happens within the world, along with more convenient ways to build and share.

Epic evolution! Google Genie 4 world model opens for experience: supports changing weather and summoning characters while playing, everything you see is interactive

Now, Google is rolling out a new Project Genie experience to eligible Google AI Ultra subscribers, greatly enhancing the responsiveness, creativity, and social attributes of these virtual worlds.

This update opens access to Genie 4 for the first time, bringing a new upgrade to Genie 3, a Project Genie agent that helps users realize their ideas, and a community gallery for discovering, exploring, and remixing worlds.

Project Genie integrates Google's most advanced world models with the Gemini family of models, allowing users to generate, enter, and shape various virtual worlds directly in the browser.

Genie 4 brings a brand-new world simulation experience: As Google's most realistic and interactive world model to date, Genie 4 supports users entering prompts mid-session to modify an already-created world.

By entering new prompts directly during navigation, users can edit the environment in real time — switching weather, adding characters, and changing the direction of action are all possible. At the same time, Genie 4 pairs visual content with real-time native audio for the first time, dynamically adapting as the scene progresses, covering everything from ambient sound effects to character dialogue.

Genie 3 receives multiple capability upgrades: Genie 4 offers the highest-fidelity visuals, physics, and native audio, while Genie 3 remains in the model selector, suited to scenarios requiring longer generation durations and generating spaces based on panoramic images.

This Genie 3 update includes: extended session duration and world chaining, increasing explorable time per generation from 60 seconds to up to 3 minutes, and allowing multiple worlds to be chained into a longer journey; support for 360-degree panoramic images as a generation basis, enabling new worlds to be generated from panoramic seeds; support for real-time in-world events, allowing the environment to be dynamically changed via natural language prompts during exploration.

Project Genie 3 can chain different worlds such as forests, underwater ruins, and frozen seas into a continuous flight journey through real-time triggerable events; it can also use 360-degree panoramic grounding to maintain the consistency of Petra Canyon and ancient building facades as the viewpoint rotates; it can further combine long-duration generation, panoramic foundations, and real-time event features to change the world at any time during flight, or even summon a dragon as a mount.

A new intelligent agent lowers the barrier to creation: The new Project Genie agent expands creative capability beyond single text or image prompts, making it as easy to build an interactive universe as having a conversation. Users simply describe their ideas in natural language, and the agent orchestrates Google's multimodal model suite to generate the initial visuals that establish the world's style, along with custom background music that matches the atmosphere.

During exploration, users can also add the Gemini model to the session: it observes what is happening on screen, pops up context-relevant suggested options, or automatically triggers in-world events as the journey progresses. For example, it can turn a natural language prompt into an interactive spacewalk, summon an alien spacecraft on command, or transform a hillside café into a desert glass pavilion. After creation is complete, users can publish custom worlds to the gallery for others to explore and remix.

The new gallery brings rich exploration content: When users simply want to jump straight into exploration, the new Project Genie gallery brings together a large collection of curated worlds and community creations. Users can turn personal photos into explorable spaces in "Living Memory," roam 360-degree panorama-based prehistoric habitats in "Dino Field Guide," or have Gemini serve as a real-time referee in "Rock Paper Scissors," judging battles between themed items on the table.

The gallery also showcases works created in collaboration with pioneering creators who gained early access to the tools: "Villa Giallo," created with creative technologist Will (aka BLVCKL!GHT), uses Genie 4 to transform a Gothic manor into an interactive mystery space that advances room by room; "The Life of Ants," created with filmmaker Anand Gandhi, leverages Genie 4's photorealism to let users explore a complex ant colony from an ant's perspective.

Beyond exploring curated worlds, users can also launch their own creations with simple prompts, or remix existing gallery works together with the Project Genie agent — adjusting the underlying world structure, soundtrack, or rules to create exclusive content. Once creation is complete, it can be published to the community gallery for others to explore and remix.

Google says it is committed to developing and promoting interactive world models responsibly. All videos and audio generated in Project Genie carry SynthID watermarks, and all model inputs and outputs undergo safety screening.

Genie 3 and Genie 4 are currently world models still in the frontier research stage. As research continues, Google is actively addressing core challenges such as camera latency, complex physical interactions, real-time audio accuracy, and long-term scene persistence.

Google Genie 4 official promotional image

Starting today, eligible Google AI Ultra subscribers can experience the new Project Genie, and Google AI Pro members can join the waitlist on the same page.