🔍 Read the full analysis: Google’s Genie 4: What’s Possible In Its Walkable AI Worlds? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google DeepMind presented Genie 4 on October 27, 2026, describing a real-time AI model that generates interactive worlds from prompts and inputs. Its selected demos show walkable scenes, generated audio and live changes, but the model is not generally available, independent performance benchmarks were not published, and scenes can drift during sessions.
Google DeepMind presented Genie 4 on October 27, describing a real-time AI model that generates interactive, walkable scenes in response to keyboard and mouse controls and typed instructions. The announcement matters because it expands the company’s world-model demonstrations to include sound generated alongside visuals and scene changes made while a world is running, though access remains restricted and the demonstrations do not establish how reliably the system works beyond selected clips.
In DeepMind’s examples, a user can move through an AI-generated scene and issue directions such as changing the weather or adding an object. The company says the system creates the next frames in response to the scene and the user’s input, rather than drawing from a conventional game engine with a stored three-dimensional map. That approach can produce scenes that appear detailed, but it can also make continuity harder to maintain as interaction continues.
DeepMind says Genie 4 generates ambient sound, music and character dialogue along with the visual scene. It also describes more natural-looking people and simulations of physical events, including collisions and aerodynamics. These are claims by the developer, illustrated with its own selected demos; DeepMind did not publish benchmark results in the source material, so the examples do not independently quantify the model’s performance.
The company says Gemini 3.8 Flash can act as a guiding agent, interpreting goals, splitting them into steps and monitoring the generated frames. A robotics-related demonstration showed a simulated Apptronik Apollo humanoid playing checkers. The source material clarifies that no physical robot took part in that demonstration: it was a simulation, not evidence of a robot completing the task in the real world.
DeepMind says Genie 4 is being tested by trusted testers and is expected to come to Google AI Ultra subscribers through Project Genie and to early Gemini API partners. It has not supplied a public launch date or an API price. Google AI Ultra is described in the source as a $200-a-month subscription tier, but that does not mean Genie 4 is currently available to all subscribers.
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
Where Interactive Worlds Could Be Used
Genie 4 points to a possible change in how teams make and test digital environments: instead of building every scene by hand, users may be able to describe a setting and explore a generated version in real time. DeepMind says testers have used the technology for film previsualization and game prototypes. Those are reported examples of early feedback, not proof that the model is ready to replace established production tools or deliver dependable results at commercial scale.
For robotics research, simulated worlds can offer a way to examine tasks and conditions without immediately relying on physical hardware. But the checkers demo establishes only that a robot was represented in a simulation. It does not show that Genie 4 has trained a real robot or improved its performance. The source says Waymo already builds its driving simulator on Genie 3, a predecessor, which provides relevant context but is not evidence that Waymo has adopted Genie 4.
Generated scenes with matching audio may also complicate how people judge whether a clip depicts a real event. DeepMind says every Genie 4 video and audio output carries an invisible SynthID watermark. The announcement does not explain how viewers will check for that mark or how well it will remain detectable after editing or redistribution.
Products Worth Considering
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
As an affiliate, we earn on qualifying purchases.
From Short Clips to Live Worlds
Genie 4 follows a series of DeepMind world-model releases. The original Genie, announced in February 2024, generated simple two-dimensional worlds; its research paper described an 11-billion-parameter model trained on unlabeled internet videos. Genie 2, introduced in December 2024, could create three-dimensional worlds from a single image and maintain them for up to a minute, though its real-time version was a reduced-quality system.
In August 2025, DeepMind described Genie 3 as its first real-time world model, running at 720p and 24 frames per second and remaining consistent for a few minutes. Google later made Genie 3 available to paying US subscribers through Project Genie in January 2026. TechCrunch reported a 60-second limit per world, partly linked to computing costs. That history helps explain why Genie 4’s reported duration and continuity matter, but it does not show that the new model has solved those limitations.
A world model differs from a finished video generator because it attempts to produce each next moment in response to what has already happened and to new user input. It also differs from a typical game engine, which uses a designed map to render a scene. In DeepMind’s account, Genie 4 predicts what comes next without a stored map, a method that can enable flexible generation but may allow objects or details to shift over time.
““a few minutes of continuous interaction””
— Google DeepMind
Limits Behind the Demonstrations
Genie 4 is not yet generally available, and the examples described are selected by DeepMind. The source reports no independent testing or benchmark figures that would show how performance varies across prompts, scenes or extended use. It is therefore unclear how consistently users can reproduce the results shown in the announcement.
The company says details and objects can drift over time and describes interaction lasting a few minutes. TechCrunch and Bloomberg, as cited by heise, reportedly put some demonstrated scenes at about two minutes, but that figure is not in DeepMind’s own announcement as described in the source. The duration a typical user can expect, and whether it differs by access tier or computing resources, remains unclear.
Other open questions include when Project Genie access will begin, which API partners will be included, and what the API will cost. The source also does not provide technical detail on SynthID detection, how generated audio and visuals will be labeled in practice, or whether the model’s simulated physics will transfer usefully to real-world tasks.
Access and Testing to Come
The next stated step is access for trusted testers, followed by planned availability for Google AI Ultra subscribers through Project Genie and early Gemini API partners. No timetable has been announced, and the source does not give a price for API use. Until those access plans become concrete, the public evidence remains DeepMind’s demonstrations and its description of the system’s capabilities.
Further testing will be needed to establish how long scenes remain coherent, how well audio and visuals respond to changing instructions, and whether the system is useful outside curated examples. Details on access conditions, session limits, pricing and independent evaluations would clarify what users and developers can actually do with Genie 4. For now, it is a research and product announcement with promising demonstrations, not a broadly available tool with independently established performance.
Key Questions
What is Google Genie 4?
Genie 4 is a world model that DeepMind says generates interactive scenes as users move through them and provide instructions. It predicts what happens next rather than playing a finished video, according to the company’s description.
Can people use Genie 4 now?
Not generally. DeepMind says trusted testers have access, with planned availability for Google AI Ultra subscribers through Project Genie and early Gemini API partners. No public launch date has been given.
Does Genie 4 generate sound?
DeepMind says it generates audio alongside visuals, including ambient sounds, music and character dialogue. This is the company’s description; the source provides no independent test results.
How long can a Genie 4 world last?
DeepMind describes support for a few minutes of continuous interaction and warns that details or objects can drift. Reports cited in the source put some scenes at about two minutes, but that figure is not stated in DeepMind’s own announcement.
Did Genie 4 show a real robot playing checkers?
No. The checkers example was a simulation of an Apptronik Apollo humanoid. The source says no physical robot was involved, so the demo does not establish real-world robot performance.
Source: ThorstenMeyerAI.com
