Google DeepMind unveils Genie 4 world model with audio and 2-minute interactive simulations

Reviewed byNidhi Govil

21 Sources

Share

Google DeepMind released Genie 4, its latest world model that generates photorealistic interactive worlds with native audio capabilities. The AI world model maintains scene consistency for up to two minutes and integrates Gemini 3.8 Flash for real-time steering, enabling users to modify simulations on the fly for robotics training, film pre-visualization, and game prototyping.

Google DeepMind launches Genie 4 with native audio and extended simulations

Google DeepMind released Genie 4 on October 27 as a research preview to trusted testers, marking a significant advancement in AI world model technology

1

2

. The latest iteration builds photorealistic interactive simulations that now include native audio capabilities, ambient sounds, music, and adaptive dialogue that responds to changing environments

4

. Unlike traditional video games requiring pre-rendered environments and programmed elements, world models generate experiences frame by frame in real time based on user prompts

1

.

Source: DeepMind

Source: DeepMind

The model maintains scene consistency for up to two minutes, double Genie 3's one-minute limit

3

4

. This extended duration allows for more complex scenarios, though DeepMind acknowledges that details and object permanence can drift during longer sessions or after multiple prompt edits

4

. According to Google, world models represent a crucial element in the march toward artificial general intelligence by creating virtual environments without traditional programming constraints

1

.

Gemini 3.8 Flash integration enables real-time generation and steering

Genie 4 leverages Gemini 3.8 Flash as a guiding agent that breaks down goals into smaller steps and evaluates each output frame

1

4

. This integration keeps simulations on track as users navigate through varying environments while maintaining original context

1

. Users can modify simulations mid-stream through natural language prompts, changing weather conditions, camera angles, or adding new people and objects

2

5

.

Source: Gadgets 360

Source: Gadgets 360

Gemini Omni anchors how Genie 4's worlds look and behave, grounding the simulations in real-world appearance

4

5

. The Gemini agent continuously observes scenes and adjusts steps in real time when users request unplanned changes

5

. Tejas Kulkarni, research director at Google, explained during a briefing that future iterations will increasingly feature hybrid world models combining Gemini writing code and text while Genie ingests that information to simulate reality at different levels of abstraction

2

4

.

Improved physics and realistic human movement expand use cases

Genie 4 delivers substantial improvements in physics simulation, handling elements like aerodynamics and collisions with greater accuracy

1

5

. Characters now exhibit more natural movement and respond organically to their environment, reducing the uncanny quality that plagued earlier versions

1

5

. These enhancements make the photorealistic simulation more useful for practical applications and visually appealing

1

.

Source: Ars Technica

Source: Ars Technica

DeepMind positions this technology for robotics training by creating infinite training environments for physical robots

1

. In one experiment tied to Gemini Robotics, a Gemini 4 Argon model broke down tasks into steps that Genie 4 simulated in real time, demonstrating how robot scenarios could be tested in simulation before real-world deployment

2

4

. The example showed Apptronik's Apollo humanoid playing checkers as a proof of concept

4

.

Film pre-visualization and game prototyping emerge as early applications

Trusted testers across film, journalism, education, gaming, and engineering have explored Genie 4 for film pre-visualization and game prototyping

4

. Diego Rivas, a Google DeepMind engineer, noted during a press briefing that the real-time aspect becomes most valuable where maintaining creative flow matters, particularly during ideation and storyboarding phases when rapid ideation is needed

2

.

Photojournalists have experimented with turning archival photos into interactive simulations

4

. One speculative example for interactive learning involves entering a painting of a historical scene and conversing with the painting's subjects about the portrayed event

2

. Rivas added that users could eventually rely on Gemini Omni to transform these prototypes into creative assets, though those capabilities remain in development

2

.

Limited access through Google AI Ultra subscription at $100 monthly

Genie 4 will roll out to developers via the Gemini API and Google AI Studio, though pricing details aren't yet available

1

. Consumer access will deploy in phases through Google AI Ultra plans starting at $100 per month

1

4

. Google may eventually expand access to AI Pro and AI Plus plans, though no specific timeline exists

1

. Access will be handled through Project Genie, which limits functionality, and a waitlist is open for users seeking early access

1

3

.

The release arrives as the world model industry consolidates. AMD agreed to acquire Fei-Fei Li's World Labs, Amazon entered talks to buy Decart, and Rivian acquired Third Dimension for self-driving simulation capabilities

2

4

. The announcement of Genie 3 last year sent gaming stocks tumbling, reflecting industry concerns about AI's potential to disrupt traditional game development

1

4

. All video and audio generated by Genie 4 carries SynthID, Google's invisible watermark for AI-generated content

4

.

© 2026 TheOutpost.AI All rights reserved