Google DeepMind has introduced Genie 4, a cutting-edge world model that generates photorealistic, explorable virtual environments in real time. Released as a research preview, the updated model adds native audio, improved physics, and deeper Gemini integration to let users steer simulations dynamically using text prompts and keyboard controls.
Artificial intelligence continues its rapid shift from static images to dynamic universes as Google DeepMind released its newest world model in research preview. Genie 4 arrives after its predecessor, Genie 3, sent shockwaves through the video game industry by generating playable virtual spaces entirely on the fly and rattling gaming stocks.
Unlike traditional virtual environments that require exhaustive manual coding or pre-rendered graphic assets, a world model creates reality frame by frame as the user explores it. DeepMind views this technology as a foundational stepping stone toward artificial general intelligence, or AGI, while commercial labs explore applications ranging from interactive learning and robotics to creative prototyping.
Real-Time Steering, Native Audio, and Realistic Physics in Genie 4
The latest iteration brings significant leaps over previous models by pairing visual generation with synchronized sensory details. Genie 4 is DeepMind’s first real-time world model to generate native audio, producing ambient soundscapes like rustling wind or crashing waves alongside interactive dialogue that adapts to user input on the fly.
Users can actively shape simulations using standard keyboard controls for navigation alongside natural language instructions entered mid-simulation. Typing prompts allows creators to instantly alter camera angles, modify the weather, or introduce new characters and objects without breaking the scene’s visual coherence. According to Google DeepMind staffers Raia Hadsell, Diego Rivas, and Hang Qi, the most dramatic leap lies in character simulation.
The most dramatic leap is how Genie 4 simulates people, creating characters that move naturally, respond organically to their environments, and can be interactively steered in real-time.
Raia Hadsell, Diego Rivas, and Hang Qi, Google DeepMind
The model maintains a consistent scene for up to two minutes while offering vastly improved handling of complex physical phenomena such as gravity and collisions.
This is Genie 4, our most realistic and interactive world model yet. 🌐Explore dynamic environments in real time to see how the scene responds, sounds, and how the people in it move, react, and speak. pic.twitter.com/6mmSUm9BBO
Gemini Integration Keeps Simulations on Track
Under the hood, Genie 4 relies heavily on Google’s Gemini ecosystem to bridge raw visual rendering with complex cognitive reasoning. When a user prompts a world into existence, Gemini 3.8 Flash operates as a guiding agent that breaks down goals, evaluates incoming frames, and verifies that each simulation step succeeds.
This oversight prevents the environment from drifting away from its original context during multi-step interactions. As Google DeepMind researchers explained, Gemini continuously monitors the scene and adapts its plan in the moment whenever the user intervenes or the world shifts.
Excited to share Genie 4 today our most realistic and interactive world model:- steer dynamic environments in real time, shaping events, camera angles, and actions- generate native audio matched to the world, from realistic background sounds to characters who can converse in… pic.twitter.com/BFgQQH9KUb
In a specialized research experiment tied to Gemini Robotics, a model variant named Gemini 4 Argon broke down high-level tasks into sequences of instructions that Genie 4 simulated in real time. Engineers envision using this pipeline to generate synthetic footage—such as a robotic arm soldering car parts—to train autonomous physical machines in simulation before deploying them in the real world.
Creative Prototyping Uses and Current Technical Limitations
Diego Rivas noted during a press briefing that trusted testers are using the technology for rapid ideation and storyboarding.
What if you could not only watch a generated video, but explore it too? 🌐Genie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt.From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵 pic.twitter.com/P0cwFvf5d2
The real time aspect becomes most valuable in the areas where keeping a state of flow in the creativity space becomes incredibly useful.
Diego Rivas, Google DeepMind engineer
Tejas Kulkarni, research director at Google, emphasized during the briefing that the future of the medium points toward hybrid architectures combining text generation with real-time simulation.
What we’ll increasingly see in the future are these hybrid world models, where you have a combination of Gemini writing code and text, and Genie ingesting that in real time and simulating reality at different levels of abstraction.
Tejas Kulkarni, research director at Google
Despite these advances, the DeepMind team acknowledges that Genie 4 remains in its early stages. Technical gaps persist: objects and details tend to drift if a simulation runs past its two-minute memory window, and engineers have not yet fully resolved how to guarantee complete audio accuracy or render legible in-world text.
Phased Rollout Plans for Developers and Subscribers
Google DeepMind is making Genie 4 available immediately as a research preview for trusted testers to gather safety and performance feedback. Broader availability is scheduled to roll out over the coming weeks.
Access will expand to Google AI Ultra subscribers participating in Project Genie, alongside early access partners utilizing the Gemini API and Google AI Studio.
Our new AI model Genie can create playable worlds in the style of 2D platformers – all from a single image prompt, sketch or text description. ✏️As a foundation world model, Genie could also help us train AI agents.Here's how. ↓ https://t.co/UX6oEkhkgi
Related reading
New Study Finds Paper Straws Have Higher Environmental Impact Than Plastic
South Korea's Ministry of Climate, Energy and Environment Upends Environmental Policy “The government must acknowledge the failure of...
Israel Election Exit Polls Show Deadlock Between Netanyahu and Eisenkot
Israel Faces Deep Political Stalemate as Exit Polls Show Deadlock Israel's wartime election ended in a tight deadlock...
Thomas Markle Sr. Probed Over Child Abuse Image Allegations
Authorities in the Philippines are investigating Thomas Markle Sr., the estranged father of Meghan, Duchess of Sussex, following...
Matthew Perry battled addiction while rising to TV fame
Netflix docuseries , debuting on October 27, 2026, explores the late star's life through rare home archives and...