{
    "@context": "https://schema.org",
    "@type": "NewsArticle",
    "@id": "https://badiyaa.com/genie-4-google-deepmind-world-model/#article",
    "mainEntityOfPage": "https://badiyaa.com/genie-4-google-deepmind-world-model/",
    "headline": "Google DeepMind&#8217;s Genie 4 world model is now available, letting Ultra subscribers rewrite AI worlds mid-",
    "description": "Google DeepMind's Genie 4 world model is now in Project Genie for Google AI Ultra subscribers, adding live audio and real-time prompt edits.",
    "datePublished": "2026-11-19T08:48:00+00:00",
    "dateModified": "2026-11-19T08:48:00+00:00",
    "author": {
        "@type": "Organization",
        "name": "Badiyaa Editorial Desk",
        "url": "https://badiyaa.com/author/editorial-desk/"
    },
    "publisher": {
        "@type": "NewsMediaOrganization",
        "@id": "https://badiyaa.com/#organization",
        "name": "Badiyaa",
        "url": "https://badiyaa.com/",
        "logo": {
            "@type": "ImageObject",
            "url": "https://badiyaa.com/wp-content/uploads/2026/10/badiyaa-logo-color-600x60-1.png",
            "width": 600,
            "height": 60
        },
        "email": "badiyaavideo@gmail.com",
        "publishingPrinciples": "https://badiyaa.com/editorial-policy/",
        "correctionsPolicy": "https://badiyaa.com/corrections/",
        "actionableFeedbackPolicy": "https://badiyaa.com/contact/",
        "ownershipFundingInfo": "https://badiyaa.com/about/#ownership",
        "sameAs": [
            "https://www.facebook.com/profile.php?id=61595518283625",
            "https://www.instagram.com/badiyaa.com.ai/",
            "https://www.threads.net/@badiyaa.com.ai"
        ],
        "parentOrganization": {
            "@type": "Organization",
            "name": "Rezilient Digital",
            "url": "https://www.rezilientdigital.com/",
            "address": {
                "@type": "PostalAddress",
                "streetAddress": "Fl No 3, Nav Shyam Sadan, Plot No 38, Pestom Sagar, Chembur",
                "addressLocality": "Mumbai",
                "addressRegion": "Maharashtra",
                "postalCode": "400089",
                "addressCountry": "IN"
            },
            "sameAs": [
                "https://www.linkedin.com/company/rezilientdigital"
            ]
        }
    },
    "isAccessibleForFree": true,
    "inLanguage": "en-US",
    "wordCount": 652,
    "image": [
        "https://badiyaa.com/wp-content/uploads/2026/11/petra-detail-of-al-khazneh-the-treasury-1783-scaled.webp",
        "https://badiyaa.com/wp-content/uploads/2026/11/petra-detail-of-al-khazneh-the-treasury-1783-1024x682.webp"
    ],
    "articleSection": "Tech",
    "keywords": "Genie 4, Google AI Ultra, Google DeepMind, Project Genie",
    "articleBody": "Google DeepMind&#8217;s Genie 4 can build a world from a sentence and then let you rewrite it while you walk through it. Type a new prompt mid-session and the weather changes, a character appears or the action swerves, and for the first time the scene comes with sound generated live, from crunching leaves to characters who talk back. The world model went live on November 18 inside Project Genie, Google&#8217;s experimental browser app, for eligible Google AI Ultra subscribers. Google AI Pro members can join a waitlist. What Genie 4 does Genie 4 is a &#8220;world model&#8221;: instead of producing a fixed video clip, it generates an explorable scene frame by frame as you move through it with a keyboard and mouse. Google DeepMind calls it its most realistic and interactive world model so far, and says it leans on the Gemini family under the hood. Gemini Omni keeps the look and behavior of a scene grounded in reality, while other Gemini models handle the reasoning that keeps a simulation consistent as it runs. On its Genie 4 page, the lab shows scenes where objects collide, water flows and trees bend in the wind, and generated characters move naturally. The showcase prompts range from a crowned crane crossing snow to a first-person dragon race and a 1980s Japanese train platform packed with commuters. Genie 4 versus Genie 3 The update also refreshes Genie 3, which stays in the model picker. Here is how the two now compare, based on Google&#8217;s own descriptions: Session length: Genie 4 supports around one minute of continuous interaction; Genie 3 now runs up to three minutes per generation, up from 60 seconds, and can chain worlds together. Sound: Genie 4 adds native, real-time audio including dialogue; Genie 3 does not get that. Starting point: Genie 3 can now be seeded from a 360-degree panorama photo; Genie 4 starts from a text description or image. Editing on the fly: both can change the world with live text prompts. The tradeoff is clear. Genie 4 gives the best visuals, physics and audio, but for a third of the time Genie 3 now allows. Someone exploring for 10 minutes would need about 10 separate Genie 4 runs, against four chained Genie 3 runs. The agent and the gallery Project Genie also gains an agent. Users describe what they want in plain language, and the agent coordinates Google&#8217;s other models to design the opening look of a world and compose background music for it. Gemini can ride along during a session, watching the screen, suggesting where to go next or firing off in-world events. A new community gallery holds curated and user-made worlds that others can explore and remix. Google highlights &#8220;Villa Giallo,&#8221; a room-by-room mystery in a gothic manor made with creative technologist BLVCKL!GHT, and &#8220;The Life of Ants,&#8221; built with filmmaker Anand Gandhi, which puts players inside an ant colony. Limits Google admits Google DeepMind is unusually blunt about the rough edges. It lists object permanence that drifts over longer sessions or after many edits, trouble with several separate characters acting at once, inaccurate recreations of real places, text that is legible only when it was in the prompt, and audio accuracy that is still being worked on. All video and audio carries SynthID, Google&#8217;s invisible watermark for AI-made content. The lab also pitches the model as a training ground. It says Gemini 4 Argon can write a multi-step robot task plan and feed it to Genie 4 as captions while the scene renders, a setup it describes as an endless playground for robotics simulation. That fits a busy month for the company, which launched its faster Gemini 4 Hydrogen model and opened pre-orders for Gemini glasses with a recording light, while DeepMind&#8217;s AlphaProof Code verified Signal&#8217;s encryption. For now, Genie 4 runs only in Project Genie for Ultra subscribers, with Pro members on the waitlist on the same page. Image: Dosseman, CC BY-SA 4.0 via Wikimedia Commons This article was produced with AI assistance and reviewed by our Editorial Desk. How we work.",
    "markdown": "https://badiyaa.com/genie-4-google-deepmind-world-model.md"
}