Genie 4: Google's World Model Explained - vs Video Generators & How to Access
Genie 4 is Google DeepMind's real-time world model with native audio. What it does, how it differs from AI video generators like Veo, and how to get access.
On October 27, 2026, Google DeepMind announced Genie 4, the newest version of its world model. The demos look a lot like AI video: photoreal landscapes, people walking and talking, wind and waves you can hear. But Genie 4 isn't a video generator, and it isn't open to the public yet. For now it's a research preview for a small group of trusted testers.
This post covers what Genie 4 actually is, how it differs from video models like Veo and Gemini Omni, how to get on the waitlist, and what you can make today while you wait.
What Is Genie 4?
Genie 4 is a real-time interactive world model. Instead of rendering a finished clip from a prompt, it generates a world you can move through as it's being created. In Google's demos, users navigate with standard keyboard and mouse controls and type text prompts to change the scene while it runs.
According to Google DeepMind's announcement, you can "set an event in motion, and Genie 4 simulates what happens next: how the world responds, how it sounds, and how the people in it move, react, and speak."
Google released its predecessor, Genie 3, as a research preview a little over a year earlier. Genie 4 adds native audio, more photoreal worlds and much tighter integration with Gemini.
What's New in Genie 4
DeepMind groups the upgrades into four capabilities:
- Real-time steering with text. You can change the weather, move the camera or add a new person or object mid-simulation. Gemini adapts your prompt to fit the scene, and between your inputs it keeps the world progressing coherently.
- More realistic physics and people. Google points to aerodynamics and collisions, and calls human simulation "the most dramatic leap": characters move naturally, react to their surroundings and can be steered in real time.
- Native audio. Genie 4 is Google's first real-time world model to generate audio natively - ambient sound, music and interactive dialogue, where characters talk and react to what you do.
- Gemini as a director. Gemini 3.8 Flash runs in the loop: it watches the scene, breaks goals into steps and checks incoming frames to confirm each step worked. Google is also experimenting with Gemini 4 Argon and Genie 4 for robotics, including a simulation of Apptronik's Apollo humanoid playing checkers.
There's also a foundation change. Per the Genie model page, Gemini Omni anchors the simulations in real-world appearance and behavior, while other Gemini models handle the reasoning that keeps a simulation coherent. That's worth remembering for later: Gemini Omni is also a video model you can use right now.
TechCrunch reports that Genie 4 can keep a scene consistent for up to two minutes, double what Genie 3 managed.
Genie 4 vs AI Video Generators: What's the Difference?
This is where most of the confusion comes from. Genie 4 and a video model like Veo 3.1 or Gemini Omni can both produce photoreal footage with sound. They solve different problems.
A video generator takes a prompt (text, an image, sometimes a video) and returns a finished clip. You wait, you get a file, you download it and post it.
A world model generates an environment frame by frame in response to your input. You're inside it while it's created. There's no fixed timeline - you decide where to look and what happens next.
| Genie 4 (world model) | AI video generators (Veo, Gemini Omni) | |
|---|---|---|
| What you get | A live, explorable world | A finished video clip |
| How you control it | Keyboard and mouse to move, text prompts to change the scene as it runs | A prompt and settings before you hit generate |
| Length | A few minutes of continuous interaction (up to ~2 min of consistent scene per TechCrunch) | Short clips, usually seconds long |
| Audio | Generated live with the world, including interactive dialogue | Generated with the clip |
| Output you can post | No public export workflow announced yet | A downloadable video file |
| Availability | Research preview for trusted testers; waitlist open | Available now |
Google itself draws this line. In TechCrunch's report, Google DeepMind engineer Diego Rivas said users could eventually use Gemini Omni to turn Genie prototypes into finished creative assets - but added that those capabilities are "still a ways off." For now, Genie 4 is a place to explore and prototype ideas, and video models are where finished clips come from.
What Genie 4 Is Good For
Google has been testing Genie 4 with trusted testers in creative industries, journalism, education, gaming and engineering. The use cases they highlighted:
- Film pre-visualization. Filmmakers use it as a "draft mode": step into a live scene and run multiple takes in real time, adjusting camera angles, lighting, dialogue and action on the fly.
- Explorable archival photos. Photojournalists are testing how a single 2D archival photo can become a grounded simulation of that place in seconds.
- Game and experience prototyping. Developers go from an idea to an explorable prototype, testing visual styles and mechanics with natural language before writing any engine code.
- Simulation. Enterprise partners are evaluating it for quick scenario generation, and Google is exploring robotics training environments.
Rivas summed up where real-time matters most: "in the ideation phase, in the storyboarding phase - anytime you're kind of in need of rapid ideation."
Genie 4 Limitations
Google is upfront that Genie 4 is "an early, experimental model." Its official limitations:
- Drift. Visual details and object permanence can shift over time, especially in longer sessions, after several complex prompt edits, or while moving around a scene.
- Short sessions. The model supports "a few minutes of continuous interaction, rather than extended sessions."
- Text. Legible text usually only appears when it's in the original world description.
- Real places. It can't simulate real-world locations with perfect accuracy.
- Multiple agents. Complex interactions between users, independent characters and the environment are still an open problem.
- Audio accuracy. Real-time audio accuracy is still being improved.
All video and audio from Genie 4 carries SynthID, Google's invisible watermark for AI-generated content.
How to Get Access to Genie 4
Right now, Genie 4 is a research preview for trusted testers only. Google says it's coming soon (TechCrunch says "in the coming weeks") to:
- Google AI Ultra subscribers in Project Genie, Google's app for generating interactive worlds from text or image prompts
- Early access partners of the Gemini API and Google AI Studio
To get in line, join the waitlist through the form linked on Google DeepMind's announcement. Google hasn't announced public API access, pricing or a general release date.
What You Can Make Today While You Wait
Genie 4 isn't available on Veevid or anywhere else for public use. But the things testers are using it for - previsualizing scenes, bringing a still photo to life, prototyping a look - can already be done with video models, as long as you're fine with clips instead of live worlds.
Storyboard and previsualize with text to video
Write a shot the way you'd describe it to a camera operator: location, light, camera move, action and sound. Generate it, change one variable and run it again. It's slower than steering a live world, but each take is a real clip you can cut together, share with a team or post.
Gemini Omni is a natural fit here because it's the same model family Genie 4 uses to anchor realism, and it generates native audio with every clip. Veo 3.1 is another strong option for realistic shots with sound.
Turn a photo into a moving scene with image to video
Genie 4's archival-photo use case has a simpler cousin you can use now: upload a still and describe the motion. Use image to video with Gemini Omni and prompt for the camera move and ambient sound, for example "slow push-in down the street, light rain starts, distant traffic and footsteps." You won't be able to walk around inside the photo, but you'll get a short, living version of it.
Mix references and existing footage with Gemini Omni
The Gemini Omni page takes up to 7 reference images in one prompt or an existing clip to guide motion, so you can keep characters, products or a visual style consistent across shots - the kind of continuity Genie 4 testers want from a draft mode.
Go from a one-line idea to a multi-shot video
If you're making a product video, the AI Video Agent writes a shot-by-shot storyboard from one sentence and a product photo, renders each shot and stitches the result with voiceover and subtitles. If you'd rather run generations from your own AI agent, see how to generate AI videos with your agent.
Should You Wait for Genie 4?
If you need a finished video this week, no. Genie 4 isn't public, Google hasn't announced an export workflow, and even its own team says the end-to-end product experience isn't settled.
If you're a filmmaker, game designer or educator who wants to explore spaces and ideas interactively, join the waitlist. Real-time steering, native dialogue and minutes-long sessions are a different kind of tool, and after trusted testers, the next wave of access goes to AI Ultra subscribers and API partners.
For most creators, the practical path is both: prototype ideas with video models now, and try Genie 4 when it opens up. For a broader look at the video side, see our best AI video generators of 2026 comparison.
Frequently Asked Questions
What is Genie 4?
Genie 4 is Google DeepMind's real-time interactive world model, announced October 27, 2026. It generates photoreal worlds with native audio that you can navigate with a keyboard and mouse and change with text prompts while they run.
Is Genie 4 a video generator?
No. A video generator returns a finished clip from a prompt. Genie 4 generates an explorable world in real time in response to your input. Google's video models are Veo and Gemini Omni; Genie 4 builds on Gemini Omni for realism.
Can I use Genie 4 now?
Not unless you're a trusted tester. It's coming soon to Google AI Ultra subscribers in Project Genie and to early access partners of the Gemini API and Google AI Studio. You can join the waitlist from Google DeepMind's announcement.
How long can a Genie 4 session last?
Google says the model supports a few minutes of continuous interaction. TechCrunch reports it can keep a scene consistent for up to two minutes, double Genie 3. Details can drift in longer sessions.
Does Genie 4 generate sound?
Yes. It's Google's first real-time world model with native audio, including ambient sound, music and interactive dialogue that reacts to your inputs.
What's the difference between Genie 4 and Genie 3?
Genie 4 adds native audio, more photorealistic worlds, more natural human characters, free-form real-time edits like changing the weather or adding objects, and tighter Gemini integration. TechCrunch reports it holds a consistent scene about twice as long as Genie 3.
Is Genie 4 available on Veevid?
No. Veevid offers Google's video models, including Gemini Omni and Veo 3.1, which you can use today to generate clips with native audio.