A new generation of world simulation

Genie 4

Our most realistic and interactive world model yet.

Genie 4 is our most realistic and interactive world model yet.

01Capabilities

Shape worlds in real time

Create and change worlds with natural language – live. Describe landscapes and events, move the camera, add new characters, or change the weather. Then see what happens next.

Genie 4 incorporates and builds on capabilities from the Gemini family. Gemini Omni anchors simulations in real-world appearance and behavior, while other Gemini models provide the complex reasoning needed to help maintain a simulation’s coherence and continuity.

Prompt-led worlds

Describe a landscape, an event, or an atmosphere and watch it take shape as a playable environment.

Physical interaction

Objects collide, cloth ripples, water moves, trees sway, and characters respond naturally.

Native sound

Footsteps, weather, crowds, and character voices arrive together with the generated world.

Interactive worlds

Create realistic interactive environments

Generate environments where objects collide, water flows, and trees sway in the wind. And where generated characters move – and behave – naturally.

01 / 04

A crowned crane walks across the snow beneath the clear dusk gradient sky. Natural colors in a documentary style.

Cause and effect

Direct complex events

Create worlds with plausible cause-and-effect sequences. Gemini works behind the scenes to interpret prompts, and make dynamic plans within changing worlds.

01 / 05

Follow an extravagantly dressed man walking backstage at an eccentric musical theatre, past performers, velvet curtains, costume racks, warm practical bulbs, and busy, lived-in corridors in one fluid handheld shot.

Responsive simulation

Direct every moment

Turn a single prompt into a sequence of believable interactions. Worlds understand the consequences of motion, timing, and contact.

01 / 03

Universal Robots UR5 holds a pipette and dispenses liquid into a lab plate, triggering tiny electric sparks.

Sound

Hear your worlds

Worlds come alive with natively generated audio. Leaves crunch underfoot, crowds cheer, and characters can speak with you – and each other.

01 / 06

A 1995 Hong Kong indie romantic drama on 35mm film: two young lovers ride a motorcycle across a bridge at dawn in one continuous, intimate shot.

Generated world audio included

02Applications

Future applications

Genie 4’s worlds have the potential for a wide range of real-world applications, from creating playful interactive experiences to generating worlds where agents – like SIMA – can learn new skills.

Frontier robotics research

Experimenting with Genie 4 and Gemini Robotics

01 / 01

Universal Robots UR5 holds a pipette and dispenses liquid into a lab plate, triggering tiny electric sparks.

We’re exploring early applications of Genie 4 for robotics.

In this example, Gemini 4 Argon creates a detailed multi-step plan to complete a task. The steps are shared as text captions to Genie 4 as the simulation is rendering.

This unlocks an infinite playground for robotics simulations.

Towards world simulation

The ability to explore an unlimited range of realistic environments is a key stepping stone on the path to AGI. Training and testing AI agents in truer-to-life environments brings us closer to their successful deployment in the physical world.

Research preview

Genie 4 is available as a research preview for trusted testers, and will be coming soon starting with Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio.

03Current boundaries

Limitations

Genie 4 is an early, evolving technology. There are many areas of ongoing research. We’re sharing Genie 4 in preview to learn alongside our community and safely advance the frontier of interactive simulation.

Coherent environments

Visual details and object permanence can shift or drift over time. This becomes more noticeable as a simulation runs longer, after multiple complex prompt edits, or when navigating throughout a scene.

Interaction and simulation of other agents

Accurately modeling complex interactions between users, multiple independent agents, and shared environments is an ongoing research challenge.

Accurate representation of real-world locations

Genie 4 is currently unable to simulate real-world locations with perfect accuracy.

Text rendering

Clear and legible text is often only generated when it's in the input world description.

Limited interaction duration

The model can support a few minutes of continuous interaction, rather than extended sessions.

Audio accuracy

Real-time audio accuracy remains an area of ongoing research and improvement.

Responsibility

Built for responsible exploration

Genie 4 marks a new area of frontier research that merges real-time generation and photorealism, introducing a new range of capabilities and dimensions that we want to roll out and explore in a responsible way. Genie 4 includes strict safety guardrails that are consistent with our safety policies across other models.

Before launching this preview, we evaluated Genie 4 through internal rigorous safety evaluations to prevent the generation of harmful or malicious content. Also, we have intentionally deployed it first to trusted tester partners to get early feedback, before we release the model more broadly. To maintain transparency, all video and audio generated by Genie 4 will include SynthID, our imperceptible digital watermarking technology that helps identify AI-generated content across visual and audio modalities.

Acknowledgements

Genie 4 was made possible by core contributions from Hang Qi, Diego Rivas, Aleksander Hołyński, Inbar Mosseri, Jack Parker-Holder, Shlomi Fruchter, Tejas Kulkarni, James Assiene, Philip Ball, Jakob Bauer, Eddie Cao, Lucy Chai, Harris Chan, Kleopatra Chatziprimou, Hao Chen, Yoni Choukroun, Mark Collier, Enric Corona, Kurtis David, Ariel Ephrat, Maxime Gazeau, Ruiqi Gao, Chelsee Glover-Odom, Liangke Gui, Swaroop Guntupalli, Shan Han, Tingbo Hou, Jiri Hron, Hui Huang, Drew A. Hudson, Zhouyuan Huo, Linyi Jin, Christos Kaplanis, Dhrupad Karwa, Eunyoung Kim, Congtao Kuang, Eliot Laidlaw, José Lezama, Hao Li, Shuang Li, Yuchen Liu, Hanzi Mao, Mark Matthews, Matt McGill, Kay McKinney, Ricardo Monti, Mark Murphy, Joseph Ortiz, Roni Paiss, Frank Perbet, Ben Poole, Tal Reiss, Konstantinos Rematas, Guy Scully, Jeremy Shar, Shashank Shastry, Sri S, Omer Tov, Ruben Villegas, Andrey Voynov, Hongjie Wang, Luyu Wang, Kaustubh Sridhar, Mo Samsami, Stephen Spencer, Yufei Wang, Will Whitney, Rundi Wu, Joyce (Jingjing) Xie, Jesse Yin, Shiran Zada, Serena Zhang, Xiaofan Zhang, Zhengqi Li, Zhixing Zhang, Yuan Zhong, Shuo-yiin Chang, Dmitry Lagun, Ronny Huang, Yao Su, Ankush Gupta, Genzhi Ye, Haolin Jia, Weiqin Wang, Xi Xiong, Salah Zaiem, Carol Zheng, Kristian Holsheimer.

Genie 4 also required close collaborations across partner teams, and we thank all our cross-functional partners that made this launch possible.

Thanks to Raia Hadsell, Aäron van den Oord, Josh Woodward, Steve Seitz, Carlos Hernandez, Elias Roman, Jung E. Kim, Abhishek Bapna, Eli Collins, Tulsee Doshi, Carolina Parada, Jie Tan, Koray Kavukcuoglu and Demis Hassabis for their insightful guidance and support throughout the research process.

Lastly, we thank our trusted tester and early access users that have provided feedback to help guide our research forward.

All recordings are of real-time interactions from Genie 4, where users use standard keyboard and mouse for navigation and camera controls, and text prompts to steer the scene as it evolves. Results for illustrative purposes. Sequences shortened.

This independent product experience is not affiliated with or endorsed by Google DeepMind. All referenced research media remains the property of its respective owner.