🔍 Read the full analysis: Google Genie 4: What Its AI Worlds Can Do Today on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google DeepMind announced Genie 4 on October 27, 2026, describing a real-time AI model that generates interactive scenes with sound and text-directed changes. The company has not released public access or benchmark results; it says the model is available to trusted testers and can drift during sessions lasting a few minutes.
Google DeepMind announced Genie 4 on October 27, describing an AI model that generates interactive worlds as users move through them and issue text instructions. The company says the system can produce visuals and sound together, but it is not publicly available, and DeepMind acknowledges that scenes can drift during sessions lasting a few minutes.
Unlike a conventional video generator, which creates a clip for later viewing, Genie 4 is designed to predict the next moment as a scene unfolds. Users can move with keyboard or mouse controls and type requests to change conditions or objects. DeepMind’s examples include changing the weather while a scene is running. The model generates the scene rather than rendering it from a stored 3D map, according to the company.
DeepMind describes four areas of progress: more realistic visuals and motion, native audio, live text-directed changes, and guidance from Gemini. Audio may include environmental sounds, music and dialogue. The company says Gemini 3.8 Flash can interpret a user’s goal, divide it into steps and monitor the generated frames. These descriptions are vendor claims illustrated with selected company demonstrations; DeepMind has not published benchmark results in the material provided.
Access is limited to trusted testers. DeepMind says it plans to make Genie 4 available soon to Google AI Ultra subscribers through Project Genie and to early Gemini API partners. It has not announced a release date or API pricing. DeepMind also says generated video and audio carry an invisible SynthID watermark.
Infographic 01 · How Genie grew up
From flat sketches to worlds with sound, in under three years.
Each step as Google DeepMind described it. Session limits and consistency are different measures, so read the labels.
- Feb 2024
Genie 1Simple 2D worlds11 billion parameters, learned from unlabelled internet videos - Dec 2024
Genie 23D worlds from a single imageConsistent “for up to a minute”; real time only as a slimmed-down, lower-quality version - Aug 2025
Genie 3First real-time Genie720p at 24 frames per second, consistent “for a few minutes”, memory about one minute back - Jan 2026
Project GenieUS AI Ultra subscribers can play60 seconds per world, partly because of compute costs - May 2026Project Genie goes globalIn the $200 AI Ultra tier, for users 18 and older
- 27 Oct 2026
Genie 4Sound, live steering, more realismResearch preview for trusted testers; Ultra subscribers and API partners “coming soon”
Genie 3 vs. Genie 4
As described by Google DeepMind. Genie 4 column = vendor claims unless marked.
| Feature | Genie 3 (Aug 2025) | Genie 4 (Oct 2026) |
|---|---|---|
| Sound | Genie 3: None | Genie 4: Generated with the picture: ambient noise, music, dialogue |
| Steering | Genie 3: Keys, plus “promptable world events” | Genie 4: Free-form text changes while it runs; Gemini adapts and checks them |
| Look | Genie 3: 720p, 24 fps | Genie 4: “Unprecedented realism”; no resolution or frame rate published |
| How long | Genie 3: “A few minutes”; Project Genie capped at 60 s | Genie 4: “A few minutes” (DeepMind); up to ~2 min consistent scene reported |
| Access | Genie 3: Small preview, then Project Genie | Genie 4: Trusted testers; Ultra and API partners “coming soon” |
Sources: Genie paper, arXiv, 23 Feb 2024; Genie 2 blog, 4 Dec 2024; Genie 3 blog, 5 Aug 2025; TechCrunch, 29 Jan 2026; Google, 19 May 2026; Google DeepMind, Genie 4 blog, 27 Oct 2026; Genie 4 model page (limitations, disclaimer); TechCrunch, 27 Oct 2026. Table wording: Thorsten Meyer.
Powered by Thorsten Meyer AI
From Clips to Interactive Worlds
If Genie 4 works beyond curated demonstrations, it could change how teams create and test digital environments. A filmmaker could explore a rough scene by moving through it; a game developer could test an idea before building a conventional game environment. DeepMind has described film previsualization and game prototypes as uses raised by testers, not as independently verified production results.
The technology also matters to robotics research, where simulated environments can help researchers test plans and scenarios. DeepMind showed a demonstration involving a simulated Apptronik Apollo humanoid playing checkers, with Gemini guiding steps. No physical robot was involved, so the clip shows a research simulation rather than evidence of a robot performing the task in the real world.
Generated scenes that include matching sound may also be harder to distinguish from recorded material. DeepMind says SynthID watermarking is applied to Genie 4 video and audio, but the announcement does not explain how reliably viewers or platforms can detect that watermark. For now, the balance is between an expanded creative and research toolset and clear limits on access, duration and consistency.
Popular right now
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Dyson Cyclone V10 Origin Cordless Vacuum
Dreame L40 Ultra Gen 2 Robot Vacuum and Mop
As an affiliate, we earn on qualifying purchases.
Genie’s Move Toward Real Time
Genie 4 follows earlier versions of DeepMind’s world-model research. The first Genie, announced in February 2024, generated simple 2D interactive environments. Genie 2, introduced in December 2024, generated 3D worlds from a single image, with DeepMind describing consistency for up to a minute. In August 2025, Genie 3 brought real-time generation, which DeepMind said ran at 720p and 24 frames per second for a few minutes.
Google later opened Genie 3 through Project Genie to some paying US subscribers, according to the supplied reporting. TechCrunch reported a 60-second limit per world, citing computing costs. That history helps explain why Genie 4’s claims about longer interaction and more natural scenes are notable, while also showing that duration and computing demands remain practical constraints.
A world model differs from a traditional game engine: it predicts what the scene should look like next instead of drawing views from a fixed, stored map. That can make interaction flexible, but it also creates a consistency challenge. DeepMind’s own description says details and objects can shift or drift over time.
“Details and objects can shift or drift over time.”
— Google DeepMind
Access, Consistency and Evidence
Several practical details remain unsettled. DeepMind has not supplied a public launch date, API price, independent benchmark results or a full account of the compute needed to run Genie 4. The duration of a typical session is also not fully established: DeepMind says interaction lasts a few minutes, while TechCrunch and Bloomberg, as cited by heise, reported scenes of up to about two minutes. That specific figure does not appear in DeepMind’s own description.
The demonstrations do not establish how consistently the model handles extended interactions, unfamiliar instructions or complex physical behavior. The company acknowledges drift, and its examples are selected by DeepMind. It is also unclear how widely the reported tester feedback on film and game workflows applies. The robotics demonstration is a simulation, not a test of a physical Apollo robot.
Waiting for Wider Testing
The next milestone is access for Google AI Ultra subscribers through Project Genie and for early Gemini API partners. DeepMind has said this is coming soon but has not given a date. Wider testing should provide more information about session length, scene consistency, audio quality and operating costs than the selected announcement clips can show.
Until access expands and independent results are available, Genie 4’s demonstrated capabilities should be treated as a preview of the product, not proof that it is ready for routine commercial or robotics use. The company’s next release details, along with reports from testers, will help clarify how the model performs outside its launch demonstrations.
Key Questions
What is Google Genie 4?
Genie 4 is a world model that generates a scene as a user interacts with it, rather than producing only a fixed video clip. Google DeepMind says users can move through scenes and request changes by typing.
Can the public use Genie 4 now?
No. DeepMind says access is currently limited to trusted testers. It has said broader access through Project Genie and the Gemini API is coming, but has not announced a date or API price.
What can Genie 4 generate?
DeepMind says it can generate interactive visuals and sound, including environmental audio, music and dialogue, and can respond to live text instructions. These capabilities are described by the company and shown in its selected demonstrations.
How long can a Genie 4 world run?
DeepMind describes a few minutes of continuous interaction and warns that scenes can drift. TechCrunch and Bloomberg, as cited by heise, reported scenes of up to about two minutes, but DeepMind’s own material does not give that exact figure.
Does Genie 4 control a real robot?
The cited Apollo demonstration was a simulation; no physical robot was involved. DeepMind has presented it as a research example, not a real-world robot trial.
Source: ThorstenMeyerAI.com
