[{"data":1,"prerenderedAt":14},["ShallowReactive",2],{"artikel-en-berita-google-genie-4-model-dunia-audio":3},{"slug":4,"lang":5,"title":6,"description":7,"date":8,"updated":9,"topic":10,"product":9,"cover":11,"readMin":12,"html":13},"berita-google-genie-4-model-dunia-audio","en","Google DeepMind launches Genie 4, a world model with real audio","Google DeepMind announced Genie 4, a world model that generates real-time interactive worlds with sound, music and dialogue, on October 27, 2026.","2026-10-28",null,"berita","\u002Fimages\u002Fartikel\u002Fberita-google-genie-4-model-dunia-audio.webp",3,"\u003Cp>Google DeepMind announced Genie 4 on October 27, 2026. It is a world model: you type a scene, it generates a world you can explore in real time. This version adds native audio, tighter physics and free-form steering in plain language.\u003C\u002Fp>\n\u003Ch2>1. What a world model is\u003C\u002Fh2>\n\u003Cp>A world model does not generate short video clips. It generates a world that keeps running: you move, the world responds. Genie 4 builds scenes from text or an image, then keeps simulating what happens next, including how the people inside move, react and speak.\u003C\u002Fp>\n\u003Cp>DeepMind calls this a crucial step toward AGI. The reason is simple: to train robots or AI agents safely, you need training environments that are infinitely variable and carry no real-world risk.\u003C\u002Fp>\n\u003Ch2>2. Audio generated with the world\u003C\u002Fh2>\n\u003Cp>Genie 4 is DeepMind's first real-time world model to generate audio natively. Every world carries its own ambient sound: wind in the trees, waves on a beach, background music.\u003C\u002Fp>\n\u003Cp>The interesting part is dialogue. Characters in a scene can talk to you and respond to changes in the world, all in real time. Every sound in DeepMind's demos was produced by the model itself, not recorded.\u003C\u002Fp>\n\u003Ch2>3. Real-time steering in plain language\u003C\u002Fh2>\n\u003Cp>You can change the scene while it runs. Type to change the weather, move the camera, or add new people and objects. A Gemini model adapts your instruction to the existing scene so the world stays consistent.\u003C\u002Fp>\n\u003Cp>Between your inputs, Gemini also monitors the scene to keep the world moving coherently. DeepMind built Genie 4 on the Gemini family: Gemini Omni anchors how the world looks and behaves, while other Gemini models supply the reasoning that maintains continuity.\u003C\u002Fp>\n\u003Ch2>4. How far it can go\u003C\u002Fh2>\n\u003Cp>According to TechCrunch, Genie 4 can hold a consistent scene for up to two minutes, double Genie 3's limit. Physics is tighter too, covering aerodynamics and collisions, and human characters look more realistic.\u003C\u002Fp>\n\u003Cp>It is not perfect. The DeepMind team itself admits details and objects can drift as a simulation runs, and in-world text is sometimes unreadable. In one experiment tied to Gemini Robotics, a Gemini 4 Argon model broke a robot task into steps and Genie 4 simulated them in real time, a way to test robot scenarios before running them for real.\u003C\u002Fp>\n\u003Ch2>5. When you can try it\u003C\u002Fh2>\n\u003Cp>Genie 4 is now in a research preview for a limited group of trusted testers. In the coming weeks it will open to Google AI Ultra subscribers through Project Genie, and to early access partners of the Gemini API and Google AI Studio. You can join the waitlist on the DeepMind blog.\u003C\u002Fp>\n\u003Cp>There is no date for general access yet. DeepMind says it is still defining what the end product experience looks like.\u003C\u002Fp>\n\u003Ch2>What it means for you\u003C\u002Fh2>\n\u003Cp>World models are still research tools, but the direction is clear: virtual worlds directed in plain language, with full sound. They will reach game developers, filmmakers and robot trainers before ordinary users do.\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cstrong>AI Ultra subscribers first.\u003C\u002Fstrong> Early access through Project Genie is limited to Google AI Ultra subscribers. If you already subscribe, watch your account.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Developers, take note.\u003C\u002Fstrong> The Gemini API and AI Studio will get early access. Anyone building simulations or learning tools should try it early.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>AI audio gets harder to spot.\u003C\u002Fstrong> Real-time generated character dialogue means fake voices keep getting better. Checking sources matters more.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Robots train in simulation.\u003C\u002Fstrong> The Gemini Robotics experiment shows robot training can be tested virtually first. That speeds up robotics development.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fdeepmind.google\u002Fblog\u002Fgenie-4-a-new-generation-of-world-simulation\u002F\" target=\"_blank\" rel=\"noopener\">Genie 4: A new generation of world simulation\u003C\u002Fa> (Google DeepMind)\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fblog.google\u002Finnovation-and-ai\u002Fmodels-and-research\u002Fgoogle-deepmind\u002Fintroducing-genie-4\u002F\" target=\"_blank\" rel=\"noopener\">We're announcing Genie 4, our new world model for real-time interactive simulation\u003C\u002Fa> (Google)\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Ftechcrunch.com\u002F2026\u002F10\u002F27\u002Fgoogles-genie-4-world-model-adds-audio-more-realistic-worlds-and-real-time-steering\u002F\" target=\"_blank\" rel=\"noopener\">Google's Genie 4 world model adds audio, more realistic worlds, and real-time steering\u003C\u002Fa> (TechCrunch)\u003C\u002Fli>\n\u003C\u002Ful>\n",1793543301828]