Genie 4: Explore AI Worlds in Real Time
Genie 4 is Google DeepMind's world model for real-time interactive simulation. Describe a world, walk through it, and change it with words as it runs, with native sound and dialogue. Genie 4 is in research preview for trusted testers. Try our interactive Genie Generator preview and join the waitlist to get notified when it opens.
Footage: Google DeepMind
What is Genie 4?
Genie 4 is a world model from Google DeepMind, announced on October 27, 2026. Instead of rendering a fixed video clip, it generates an interactive world you can explore in real time with your keyboard and mouse. You can steer the scene with natural language while it runs, and every world comes with its own ambient sound, music and dialogue.
How to Use Genie 4 (and Get Access)
Genie 4 is not publicly available yet. Here is where access stands and what you can do today.
Official research preview
Google DeepMind is rolling Genie 4 out as a research preview to trusted testers first, gathering feedback before a wider release.
Coming next: Project Genie and the Gemini API
Google says Genie 4 is coming to Google AI Ultra subscribers through Project Genie, and to early access partners of the Gemini API and Google AI Studio. No public date or API price has been announced.
Join the official waitlist
Google DeepMind runs its own waitlist for Genie 4 in Project Genie. The link is published on the official Genie model page. Signing up does not guarantee access.
Write a focused first prompt
Describe one place, one point of view and one mood. A tight scene is easier for a world model to keep coherent than a sprawling one.
Explore first, then change one thing
Walk around before you edit. Then steer with one change at a time, such as the weather or a new character, so you can see what each edit does.
Practice in the preview
Open the Genie Generator on this site to try an interactive demo of the workflow, then join our waitlist to hear when real Genie 4 worlds are available here.
Worlds You Can Walk Into
Genie 4 turns a short description into a living scene that responds to you, frame by frame.
Real-time interactive worlds
Genie 4 generates the world as you move through it. Walk, look around and act, and the simulation reacts in real time instead of playing back a pre-rendered video.
Steer the scene with words
Type a change while the world is running, such as switching the weather, moving the camera or adding a new person or object, and Genie 4 adapts the scene on the fly.
What's New in Genie 4
Compared with Genie 3, Google DeepMind highlights these upgrades.
Native audio
Google calls Genie 4 its first real-time world model to generate audio natively: ambient sound, music and characters that speak, matched to what happens in the world.
Coherent short sessions
Google says Genie 4 can keep environments coherent over a few minutes of continuous interaction, so places and objects hold together while you explore. It is built for short sessions, not hours-long play.
More realistic physics and people
Movement, collisions and human motion look more natural, from a falling seed spinning to the ground to characters that move like real people.
Built with Gemini
Gemini Omni anchors each world in real-world appearance and behavior, and Gemini adapts your prompts to the scene. In Google's complex-event demos, Gemini 3.8 Flash also breaks goals into steps and reviews incoming frames to check each one.
Genie 4 Prompt Examples
Starter world descriptions and a follow-up edit for each. These are our own examples written for the workflow, not outputs tested on Genie 4.
Rainy neon alley
"A narrow alley in a rainy city at night, neon signs reflected in puddles, steam rising from a vent, seen at walking height." Then try: "The rain stops and a food cart opens at the end of the alley."
Alpine meadow at dawn
"A wide alpine meadow at sunrise, wildflowers, a wooden fence and a distant lake, cowbells in the distance." Then try: "Fog rolls in from the lake."
Toy-scale kitchen
"A kitchen counter seen from the height of a toy car, giant cups and a sliding bowl of fruit, warm morning light." Then try: "Someone turns on the tap and water spills across the counter."
What Can You Use Genie 4 For?
Early directions based on Google DeepMind's demos. Treat generated events as drafts to validate, not ground truth.
Game and experience prototyping
Sketch a level, a mood or an interaction and walk through it before anyone builds assets. Genie 4 is not a game engine, so it suits ideas, not shipping builds.
Film previs and storyboarding
Block out a location, try camera moves and hear a rough soundscape to plan a shot before a shoot or a full render.
Robotics and agent research
Google DeepMind shows Gemini planning multi-step tasks that Genie 4 simulates, pointing to world models as training and testing grounds for agents.
Genie 3 vs Genie 4
Only what Google DeepMind has published. Where Google gives no figure, we say so.
| Genie 3 | Genie 4 | |
|---|---|---|
| Announced | August 5, 2025 | October 27, 2026 |
| Real-time exploration | Yes | Yes, with more realistic physics and human motion |
| Native audio | Not mentioned by Google | Yes: ambient sound, music and dialogue |
| Change the world with text while it runs | Yes, via promptable world events | Yes, with Gemini interpreting prompts and planning |
| Session length | A few minutes | A few minutes of continuous interaction |
| Resolution and frame rate | 720p at 24 fps | Not published |
| Access at launch | Limited research preview for academics and creators | Research preview for trusted testers, then Project Genie (Google AI Ultra) and the Gemini API |
| Public price | Not published | Not published |
Genie 4 Limitations
Google DeepMind calls Genie 4 an early, experimental model. These are the limits it states itself.
Drift over time
Visual details and object permanence can shift, especially in longer runs, after several complex edits, or while moving through the scene.
Physics and multiple characters
Complex physical interactions and interactions between several agents are still open research problems.
Text, places and audio
Legible on-screen text usually appears only if it is in the description, real-world places are not reproduced exactly, and real-time audio accuracy is still improving.
Watermarked output
All video and audio generated by Genie 4 carries SynthID, Google's imperceptible digital watermark.
Why Follow Genie 4 on Genie4.io?
Genie4.io is an independent hub for people who want to understand Genie 4 and be ready the day it opens up.
Clear, up-to-date coverage
Plain-language explainers on what Genie 4 can do, what is still limited, and how access is rolling out.
Hands-on preview
Try the Genie Generator: an interactive demo of the world-model workflow, with live controls and text steering. No credits used.
Get notified first
Join the waitlist with the world you want to build, and we'll let you know when Genie 4 can generate it here.
Independent and transparent
Genie4.io is not affiliated with Google or Google DeepMind. We say clearly what is a demo and what is the real model.
Frequently Asked Questions
Quick answers about Genie 4 and this site.
1. What is Genie 4?
Genie 4 is Google DeepMind's world model for real-time interactive simulation. It generates explorable worlds from a description, lets you steer them with natural language while they run, and produces matching sound, music and dialogue.
2. Who made Genie 4?
Genie 4 was built by Google DeepMind and announced on October 27, 2026. It follows Genie 3, which was announced as a limited research preview in August 2025.
3. Can I use Genie 4 right now?
Not yet for most people. Genie 4 is currently a research preview for trusted testers. Google has said it is coming soon to Google AI Ultra subscribers through Project Genie.
4. How do I get access to Genie 4?
Join Google DeepMind's official Genie 4 waitlist for Project Genie, linked from the official Genie model page, and watch for the Google AI Ultra rollout and early access through the Gemini API and Google AI Studio. You can also join our waitlist and we'll email you when Genie 4 worlds can be generated on Genie4.io.
5. How is Genie 4 different from Genie 3?
Genie 4 is Google's first real-time world model to generate audio natively, uses Gemini to interpret prompts and plan inside the world, and shows more realistic physics and human movement. Both models support a few minutes of interaction and text-driven changes. Our full Genie 4 vs Genie 3 comparison covers every published difference.
6. Is Genie 4 free? How much does it cost?
No public price or free tier has been announced. Broader access is planned through Google AI Ultra in Project Genie, but Google has not said when Ultra subscribers will get it, and no Gemini API pricing exists yet.
7. When will Genie 4 be released to everyone?
There is no public release date. Google says it is coming soon, starting with Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio.
8. Is there a Genie 4 app or download?
No. There is no standalone Genie 4 app to download. Access is planned through Project Genie on the web and through Google's developer tools. Be wary of sites offering a Genie 4 download.
9. Is Genie 4 a game engine?
No. Genie 4 generates each frame as you interact instead of running game code, and Google says it supports a few minutes of continuous interaction. It is useful for prototyping and research, not for shipping games.
10. What resolution and frame rate does Genie 4 run at?
Google has not published a resolution or frame rate for Genie 4. It says the model supports a few minutes of continuous interaction rather than extended sessions. For reference, Genie 3 was announced at 720p and 24 frames per second.
11. Is there a Genie 4 API?
Not from Genie4.io: this site does not offer an API or API keys. Google plans to give early access partners Genie 4 through the Gemini API and Google AI Studio, but no public endpoint, model ID or pricing has been published.
12. Is Genie4.io the official Genie 4 website?
No. Genie4.io is an independent site and is not affiliated with Google or Google DeepMind. Genie is a trademark of Google LLC. For official information, see the Google DeepMind website.
13. What is the Genie Generator on this site?
It is an interactive demo preview of the world-model workflow: you describe a world, move through it with WASD, and steer it with words. It does not run Genie 4 and does not use credits. It is clearly marked as a demo, and we will connect the real model when access opens.
Official Genie 4 Sources
Everything on this page is checked against Google DeepMind's own pages. Last reviewed October 29, 2026.
- Genie 4 announcement (Google DeepMind blog) ↗
Launch post covering capabilities, rollout and limitations.
- Genie model page ↗
Official overview, demos, responsibility notes and the waitlist link.
- Genie 4 launch film (YouTube) ↗
Google DeepMind's official announcement video.
Get Ready to Explore Genie 4 Worlds
Try the interactive preview today and be the first to know when Genie 4 opens.