- Google DeepMind built Genie 4 to create interactive worlds from text prompts.
- Users can alter weather, objects, and camera angles mid-simulation without restarting.
- Simulations run for up to two minutes with native audio and dialogue.
- Access starts with trusted testers and Google AI Ultra subscribers on Project Genie.
In this guide
What is Genie 4?
Genie 4 is an artificial intelligence world model created by Google DeepMind to generate interactive virtual environments from text prompts. It builds upon Genie 3 by delivering higher visual realism, natural character movement, and responsive real-world physics.
- Developer
- Google DeepMind
- Model type
- Interactive world model
- Simulation duration
- Up to 2 minutes
- Audio generation
- Native ambient sounds, music, and dialogue
- Real-time editing
- Prompt steering during live simulation
- Consumer entry point
- Project Genie via Google AI Ultra ($100/mo)
- Developer access
- Gemini API and Google AI Studio
Unlike standard video generation tools that output static clips, Genie 4 produces environments that users can actively explore. Google DeepMind released the model as a research preview to evaluate how simulated realities can support creative workflows and autonomous software testing.
What can Genie 4 do?
Genie 4 generates photorealistic simulations that respond continuously to user directions and physical interactions. It introduces native audio generation, producing ambient background noise, music, and interactive dialogue that adapt directly to events within the simulated world.
A key capability of Genie 4 is real-time steering. Users can type new prompts mid-simulation to alter the scene, such as introducing new objects, modifying weather conditions, or shifting camera perspectives without restarting the render.
How does Gemini assist Genie 4?
Genie 4 relies directly on Google's Gemini models to supervise and structure simulated worlds. Google integrates Gemini 3.8 Flash to function as an underlying director that breaks user goals into logical steps and verifies each visual frame during generation.
This integration significantly improves physical accuracy, including collision detection and aerodynamic behavior. Digital Trends reports that people generated within these worlds move with greater fluidity and less stiffness than in earlier iterations.
In robotics research, Google DeepMind paired Genie 4 with Gemini 4 Argon. The Gemini model plans robotic actions, and Genie 4 visualizes and tests those steps in real time before hardware deployment.
Who is Genie 4 designed for?
Genie 4 serves specialized technical and creative teams requiring rapid spatial prototyping. Google identifies immediate applications in previsualization for filmmaking, game design concept testing, and synthetic training for autonomous robots.
- Game studios prototyping dynamic environments and level mechanics
- Filmmakers exploring real-time storyboarding and scene staging
- Robotics teams validating control policies in physics simulations
- Long-form video productions requiring scenes past two minutes
- Projects that require guaranteed text legibility inside environments
- Teams seeking fully finalized, production-ready commercial assets
TechCrunch notes that the broader world model sector is expanding rapidly. Genie 4 arrives as competitors like Runway offer playable world models and hardware makers invest in real-time simulation technology.
How to get access to Genie 4
Genie 4 is currently restricted to a research preview for selected trusted testers. Google has provided a waitlist for external users interested in trying the system as access widens.
Consumer access will launch through Project Genie, Google's tool for building worlds from text or image prompts, beginning with AI Ultra subscribers.
Google plans to roll out consumer access to subscribers of its Google AI Ultra tier, which costs $100 per month. Broader availability on AI Pro or AI Plus subscription plans may arrive at a later stage.
Software developers will eventually access Genie 4 through the Gemini API and Google AI Studio. Google has not yet announced final API pricing or release dates for commercial developer tiers.
What are the limits of Genie 4?
Genie 4 remains an early-stage research system with several technical constraints acknowledged by Google DeepMind. The primary limitation is temporal consistency: scenes can only maintain stability for up to two minutes before visual drift occurs.
As simulations progress, the model can lose track of fine details and struggle with object permanence. Google also notes that maintaining precise audio synchronization and rendering legible text inside simulated environments remain unsolved technical challenges.
Genie 4 simulations can drift after a couple of minutes, causing objects to disappear or scene consistency to degrade.
Frequently asked questions
What is Google Genie 4?
How long can a Genie 4 simulation run?
Can you change a Genie 4 world while it is running?
How much does it cost to use Genie 4?
What are the main technical limits of Genie 4?
Sources: Digital Trends · TechCrunch
Want your brand to be the answer AI gives?
See how ready your website is for ChatGPT, Gemini and Perplexity, free, in about ten seconds.