Google DeepMind launches real-time world model Genie 4 with Gemini models planning in the background

What happened

Google DeepMind has introduced Genie 4, the successor to Genie 3, which the company showed in August 2025. Unlike a video generator, Genie 4 does not produce finished videos but calculates a scene frame by frame, reacting to input as it goes. Users move through the scene with the keyboard and steer events with text prompts. For now, only select testers can try it.

A notable architectural feature is the division of labor: Gemini models plan in the background what happens next, while the underlying generation is handled by a video model. DeepMind frames this pairing of language-model reasoning with pixel-level simulation as an early step toward hybrid world models.

Key details

• Genie 4 is DeepMind's first real-time world model that generates sound directly. Each world has its own background audio, such as wind in the trees, and characters are supposed to talk to each other and react in real time to user input or events.
• DeepMind sees the biggest gain in realism in how the model renders people: characters should move naturally, react to their surroundings, and be directly controllable by users.
• The model builds on the video model Gemini Omni so the world looks and behaves realistically; other Gemini models help plan the simulation to keep it consistent. If a user changes the weather with a text prompt, Gemini adapts the instruction to the running scene, and it also steers events between inputs so the world keeps developing in a way that makes sense.
• In a multistep-task example, Gemini 3.8 Flash watches the scene continuously, breaks a goal into steps, and uses new frames to check whether each step worked, adjusting its plan if the world changes or the user steps in.
• DeepMind is also testing Gemini 4 Argon to tie the Genie work to its robotics research, Gemini Robotics: in one demo, Argon breaks the task of playing checkers into steps, which Genie 4 receives as text instructions to simulate Apptronik's humanoid robot Apollo playing the game, with no real robot involved.
• Limitations are explicit: DeepMind calls Genie 4 an early, experimental model. Worlds stay consistent over short sessions, but visual details can shift and objects do not always persist in longer simulations or after several complex prompt changes. Complex physical interactions, accurate real-time sound, and readable text are still under research.

Why it matters

The significance of Genie 4 lies in combining a real-time interactive world model with language-model planning: Gemini keeps the world developing coherently between user inputs, moving beyond the passivity of pure generative models. Text instructions get adapted to the running scene, and multistep goals can be decomposed, verified frame by frame, and adjusted dynamically—closer to a controllable simulation environment than a video generator.

DeepMind also points to the long-term direction: once mature, world models could serve as training environments where robots and AI agents learn from simulated experience rather than just from data. The checkers demo shows a full pipeline from task decomposition to simulated robot execution, which could become a new paradigm for robotics training.

Who it affects

Access is currently limited: Genie 4 is available as a research preview to select testers, and is set to reach Google AI Ultra subscribers in the Project Genie web app soon, along with partners with early access to the Gemini API and Google AI Studio. All videos and audio tracks will carry Google's SynthID watermark. For robotics researchers and agent developers, the simulated-training-environment direction is worth watching; for creators, real-time interactive worlds with sound and dialogue could become a new content format, though limits like long-session consistency will constrain practical use for now.

Genie 4 is supposed to simulate interactive worlds in real time, including sound and talking characters. Gemini models plan in the background what happens next in the simulation. Only select testers can try the world model for now.

Google Deepmind has introduced Genie 4, the successor to Genie 3, which the company showed in August 2025. The world model doesn't produce finished videos but calculates a scene frame by frame, reacting to input as it goes. Users move through the scene with the keyboard and steer events with text prompts.

According to Deepmind, Genie 4 is the company's first real-time world model that generates sound directly. The company says each world has its own background audio, such as wind in the trees. Genie 4 also handles dialogue, with characters that are supposed to talk to each other and react in real time to user input or events.

Deepmind sees the biggest gain in realism in how the model renders people. Characters are supposed to move naturally and react to their surroundings, and users should be able to control them directly.

Gemini models plan ahead to keep the world coherent

According to Deepmind, Genie 4 builds on capabilities of the video model Gemini Omni so the world looks and behaves realistically. Other Gemini models help plan the simulation to keep it consistent.

If a user changes the weather with a text prompt, Gemini is supposed to adapt the instruction to the running scene. Deepmind says Gemini also steers events between inputs so the world keeps developing in a way that makes sense.

Deepmind explains the process with multistep tasks. Gemini 3.8 Flash watches the scene continuously, breaks a goal into steps, and uses new frames to check whether each step worked. If the world changes or the user steps in, the model adjusts its plan.

Deepmind is also testing Gemini 4 Argon to tie its Genie work to its robotics research, Gemini Robotics. In one demo, Argon breaks the task of playing checkers into steps. Genie 4 gets these as text instructions and uses them to simulate Apptronik's humanoid robot Apollo playing the game, with no real robot involved.

Longer sessions still trip up the model

Deepmind calls Genie 4 an early, experimental model. According to the company, the world stays consistent over short sessions, but visual details can shift and objects don't always persist in longer simulations or after several complex prompt changes.

Deepmind says it's still researching complex physical interactions, accurate real-time sound, and readable text.

For now, Genie 4 is only available to select testers as a research preview. It's set to reach Google AI Ultra subscribers in the Project Genie web app soon, along with partners who have early access to the Gemini API and Google AI Studio.

Google says all videos and audio tracks will carry its SynthID watermark, which invisibly marks AI-generated content.

Once they mature, world models could serve as training environments where robots and AI agents learn from simulated experience rather than just from data. Deepmind sees pairing Gemini with Genie as an early step toward hybrid world models that combine the reasoning of language and code models with pixel-level simulation.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now

The original belongs to The Decoder and its authors and is reproduced here for reading. To have it removed, write to support@aisa.one.