TL;DR: Google DeepMind has announced Genie 4, a new world model that generates explorable, photorealistic scenes in real time and, for the first time in the Genie line, creates audio natively, including ambient sound, music and interactive dialogue. You can steer a running scene with plain-language prompts, such as changing the weather or adding a character. Genie 4 is a research preview for trusted testers right now; Google says it’s coming soon to Google AI Ultra subscribers in Project Genie and to early-access partners in the Gemini API and Google AI Studio. Generated video and audio carry SynthID watermarks.
On October 27, 2026, Google DeepMind unveiled Genie 4, the next version of its “world model” research. A world model doesn’t just render a clip and stop. It simulates what happens next, frame by frame, while you move around inside it and change things as you go. Genie 4 adds two things earlier versions lacked: much more realistic people and physics, and sound that’s generated together with the picture. Here’s what Google DeepMind says the Genie 4 world model can do, how it works with Gemini, who can use it and where it still falls short.
What is Genie 4?
According to Google DeepMind’s announcement, Genie 4 is a natively multimodal world model for real-time interactive simulation. You describe an environment (or start from a reference image), and Genie 4 builds a world you can move through with standard keyboard controls for navigation and the camera. Text prompts then steer the scene while it runs.
Google DeepMind says the model “incorporates and builds on capabilities from the Gemini family.” In its description, Gemini Omni anchors the simulations in real-world appearance and behavior, while other Gemini models handle the reasoning needed to keep a world coherent from moment to moment.
If you’re new to how generative models build images, our explainers on diffusion models and multimodal AI are useful background.
The four headline capabilities
Google DeepMind groups Genie 4’s improvements into four areas. Every claim below is the company’s own description from its launch post.
| Capability | What Google DeepMind says it does |
|---|---|
| Real-time steering | Change the weather, move the camera or introduce a person or object with natural language; Gemini adapts the prompt to fit the scene. |
| Realism in physics and people | Simulates scenarios such as aerodynamics and collisions; characters move naturally and can be steered in real time. |
| Native audio | Each world carries its own ambient sound, and characters can hold interactive dialogue that reacts to your inputs. |
| Complex events with Gemini | Gemini 3.8 Flash acts as a guiding agent that plans steps, watches incoming frames and checks progress. |
Steering a world while it runs
With Genie 4 you “set an event in motion,” and the model simulates how the world responds, sounds and moves. Between your inputs, Google says Gemini also “monitors and steers the scene to keep the world progressing cohesively,” so a world doesn’t simply freeze or fall apart when you stop typing.
Sound generated with the picture
This is the biggest change from earlier Genie models. Google DeepMind calls Genie 4 its first real-time world model to generate audio natively, from wind in the trees to waves on a beach. It calls human speech “the true frontier”: characters can converse and react to you or to events in the world in real time.
Gemini as a planner inside the loop
For multi-step tasks, Google describes a setup where Gemini 3.8 Flash observes the scene as Genie 4 generates it, breaks a goal into steps and reviews new frames to confirm each step worked. If the world shifts or you intervene, Gemini adjusts the plan. The company also shows an experiment pairing Genie 4 with Gemini 4 Argon to simulate Apptronik’s Apollo humanoid robot playing checkers, with Argon feeding step-by-step captions to Genie as the simulation renders.
Who can use Genie 4, and when?
Not most people, yet. Google’s Keyword blog post says Genie 4 is rolling out to a limited group of trusted testers first. The rollout order Google gives is:
- Trusted testers (now), as a research preview used to gather feedback.
- Google AI Ultra subscribers in Project Genie (“coming soon”), with a waitlist open now.
- Early-access developer partners in the Gemini API and Google AI Studio (“coming soon”).
Google hasn’t published pricing, usage limits or a firm date for the wider release.
What testers are using it for
Google DeepMind says it has worked with tester groups in creative industries, journalism, education, gaming and engineering. The uses it highlights include:
- Pre-visualization: filmmakers stepping into a live scene and running multiple “takes,” adjusting camera angles, lighting, dialogue and action on the fly.
- Archival photos as places: photojournalists testing how a single historical photograph can become a grounded, explorable simulation, or how short events such as a flash flood unfold.
- Game prototyping: developers going from an idea to an explorable prototype, testing visual styles and mechanics through prompts before writing engine code.
- Scenario generation: enterprise partners evaluating it for quick simulations, and DeepMind’s own robotics research.
Google frames this as part of a bigger goal. In its words, “understanding and simulating the world is a crucial aspect of building AGI,” and safe, varied simulations can serve as training grounds for robots and AI agents.
Limits Google admits to
Google DeepMind is unusually direct that Genie 4 is “an early, experimental model.” The limitations it lists:
- Visual details and object permanence can drift over time, especially in longer sessions, after many complex prompt edits or while moving around the scene.
- Complex physical interactions, real-time audio accuracy, legible text, spatial accuracy and long-term consistency are still active research areas.
That matters if you’re picturing Genie 4 as a game engine replacement. Coherence over “short sessions” is the claim; a world that stays consistent for hours is not.
Safety and watermarking
Google says Genie 4 follows the same safety policies as its other models, went through internal safety evaluations before the preview and was deliberately released to trusted testers first. All video and audio Genie 4 generates will include SynthID, Google’s imperceptible watermark for identifying AI-generated content across images and audio.
Why Genie 4 matters
Most AI video tools produce a fixed clip. Genie 4 is aimed at something different: a world that keeps running and responds to you, with sound. If Google can extend how long those worlds stay consistent, the same technology could feed robot training, film pre-production and game design. For now it’s a research preview, and the most useful thing for everyone else to watch is when the Project Genie waitlist opens to Ultra subscribers and what limits come with it.
Google also announced on the same day that Gemini Robotics ER 2 is in public preview on Google Cloud, a sign that its world-model and robotics work are moving closer together. For more on how today’s AI models reason, see our guide to reasoning models.
FAQ
What is Genie 4?
Genie 4 is Google DeepMind’s latest world model. It generates interactive, real-time simulations you can move through and steer with text prompts, and it generates matching audio natively.
Can I try Genie 4 now?
Only if you’re a trusted tester. Google says it’s coming soon to Google AI Ultra subscribers through Project Genie, which has a waitlist, and to early-access partners in the Gemini API and AI Studio.
Is Genie 4 a video generator?
Not exactly. A video generator produces a finished clip; Genie 4 keeps simulating the world as you navigate and change it, frame by frame.
How can you tell Genie 4 content is AI-made?
Google says all Genie 4 video and audio includes SynthID, its invisible digital watermark.
The Teck Hustlers Editorial Team covers technology, business, finance and gaming news for readers who want the story behind the headline. Every article is researched, written and edited by our in-house team, checked against company announcements, filings and reputable reporting, and updated when new facts emerge. We focus on what each development means for founders, professionals, investors and everyday users. Have a tip or a correction? Reach us through our contact page.
