Skip to content
major New models verified update

Google has launched production access to the Genie 4 model for Google AI Ultra subscribers

confirmed by 4 independent sources updated November 18, 2026

Google has launched production access to the Genie 4 model (previously only a research preview) for Google AI Ultra subscribers in the Project Genie app, with world edits during simulation and real-time audio. It has also updated the Genie 3 model and added a community gallery.

Google has launched a gradual production rollout of the Genie 4 model for select Google AI Ultra subscribers within the Project Genie web application. The model was previously available only to select testers in a research preview mode; the company has now also updated the older Genie 3 model (sessions extended from one to up to three minutes, support for 360° panoramic images, and the ability to chain multiple worlds into longer journeys), added a new Project Genie agent for building worlds from spoken natural-language descriptions, and introduced a community gallery for discovering and remixing created environments. Google AI Pro subscribers can join the waitlist.

Genie 4 is a world model that generates interactive 3D environments frame by frame in real time. Users can modify the world by entering a new prompt in the middle of the simulation – changing the weather, adding characters, or directing the plot. According to the company, it is the first model in this series to natively generate real-time audio, including ambient sounds and character dialogue. The company states that the model can maintain a consistent scene for up to two minutes (compared to one minute for the Genie 3 model) and improves the simulation of physics and the appearance and behavior of characters; the Gemini 3.8 Flash model is used to plan the simulation steps.

Google cites training robots in simulated environments and rapid prototyping, for example in storyboarding for filmmaking, as possible use cases. All generated video and audio content carries a SynthID watermark and undergoes safety checks. The company also states that the model still has limitations – during longer simulations and more complex changes, details and objects may drift, audio may not always be accurate, and text in the scene may not be legible.

What changed

Why it matters

Making the Genie 4 model available to a wider circle of subscribers (previously only select testers) moves world models from the experimental stage towards real-world use – creators can use the tool for rapid prototyping of visual concepts and storyboards, while companies working in robotics can monitor its potential for simulated training. However, the model is still in an early stage with stated limitations around longer simulations, complex physics, and audio accuracy, so for now it is a supplementary tool rather than a finished production solution.

Release card

Genie 4

Google DeepMind

Inputs
The input is text prompts, and in the case of the Genie 3 model also 360-degree panoramic images; the output is interactive video generated frame by frame in real time with native audio including music, ambient sounds, and character dialogue.
Availability
Available in the Project Genie tool for eligible Google AI Ultra subscribers; Google AI Pro subscribers can join the waitlist.
According to the sources, it is suitable for
  • Pre-visualization and rapid storyboarding for filmmakers
  • Prototyping game ideas
  • Creating training scenarios for robots in simulation
Documented limits
  • During longer simulations, details and objects in the scene may drift
  • The accuracy of generated real-time audio has not yet been resolved
  • Text displayed within the scene may be illegible

The model distinguishes itself from its predecessor Genie 3, over which it offers higher photorealism, better physics, more human-like characters, and now native audio as well as the ability to change the scene via a prompt in the middle of the simulation.

The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →

What was added since the original report

Verified updates

  1. New verified information

    Genie 3 will be updated to support sessions of up to 3 minutes (compared to 1 minute originally); Genie 3 gains 360° panorama and chainable worlds features; The community gallery includes SynthID watermarks and safety checks; A new AI agent enables building worlds from spoken natural-language descriptions; Genie 4 enables real-time scene edits via text prompts (weather, characters, plot)

    • Genie 3 will be updated to support sessions of up to 3 minutes (compared to 1 minute originally)
    • Genie 3 gains 360° panorama and chainable worlds features
    • The community gallery includes SynthID watermarks and safety checks
    • A new AI agent enables building worlds from spoken natural-language descriptions
    • Genie 4 enables real-time scene edits via text prompts (weather, characters, plot)
  2. New verified information

    Genie 4 is available today in a gradual rollout (rather than a future announcement); The Project Genie agent orchestrates multimodal models for creation; The community gallery enables discovering, exploring, and remixing worlds; Genie 3 now supports sessions of up to 3 minutes and 360° panorama grounding; Native real-time audio adapts to the scene

    • Genie 4 is available today in a gradual rollout (rather than a future announcement)
    • The Project Genie agent orchestrates multimodal models for creation
    • The community gallery enables discovering, exploring, and remixing worlds
    • Genie 3 now supports sessions of up to 3 minutes and 360° panorama grounding
    • Native real-time audio adapts to the scene
  3. New verified information

    Access is to later be expanded to Google AI Ultra subscribers.; According to the company, the model has limitations in longer simulations and complex physics.

    • Access is to later be expanded to Google AI Ultra subscribers.
    • According to the company, the model has limitations in longer simulations and complex physics.
  4. New verified information

    The generated audio includes music, ambient sounds, and interactive dialogue.; The Gemini model plans steps and checks frames when working together with Genie 4.

    • The generated audio includes music, ambient sounds, and interactive dialogue.
    • The Gemini model plans steps and checks frames when working together with Genie 4.
  5. New verified information

    Google states an improvement in physics simulation compared to Genie 3.; Google states an improvement in the simulation of people compared to Genie 3.; The model generates environments frame by frame in real time.

    • Google states an improvement in physics simulation compared to Genie 3.
    • Google states an improvement in the simulation of people compared to Genie 3.
    • The model generates environments frame by frame in real time.

Two audiences, two different impacts

What this means

01

For individuals

Google AI Ultra subscribers now have direct access to the Genie 4 model and the updated Genie 3 model in Project Genie, including the ability to modify generated worlds with text prompts in the middle of a simulation, listen to generated audio, and share or remix creations in the community gallery.

What to do Try the new version of Project Genie as part of a Google AI Ultra subscription, or join the waitlist via Google AI Pro.
More practical updates →
02

For a business

According to Google, the Genie 4 model has potential for training robots in simulation and for rapid prototyping in filmmaking and game creation, but access is currently limited to Google AI Ultra and select Gemini API partners, so for companies this is currently a matter of monitoring developments rather than deployment.

Development
What to decide Monitor the planned expansion of access to the Genie 4 model via the Gemini API and Google AI Studio for potential use in prototyping or simulated robot training.
More business impacts →
3D generation community editor Gemini Gemini 3.8 Flash Gemini Robotics Genie 4 Google Google DeepMind interactive generation interactive worlds multimodal AI Project Genie world model world models

Check the original

Event sources

confirmed by 4 independent sources · 5 publishers, 4 independent. We count feeds from the same owner only once.

7
The Verge AI independent context · first detected Google’s next Genie AI model generates more realistic experiences Ars Technica (AI) independent context DeepMind unveils Genie 4 world model promising more realistic physics and humans Google DeepMind Blog primary source Genie 4: A new generation of world simulation TechCrunch AI independent context Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering The Decoder (daily AI news) independent context Google Deepmind's Genie 4 puts a modern spin on agent designs that predate LLMs Google DeepMind Blog primary source Genie 4 is now available The Decoder (daily AI news) independent context Google Deepmind's Project Genie update brings Genie 4, chainable worlds, and a community gallery for remixing