🔍 Read the full analysis: What’s New About Google’s Genie 4? Walkable Worlds And Their Limits on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Google DeepMind presented Genie 4 on October 27, 2026, demonstrating interactive scenes that users can explore and alter with text while the model generates visuals and sound. The demonstrations are company-selected, public access has not opened, and DeepMind says scenes can drift and support only a few minutes of interaction.
Google DeepMind demonstrated Genie 4 on October 27, presenting an AI model that generates scenes users can move through and change with text while it runs. The company says the model also generates sound alongside visuals, but access remains limited to trusted testers and DeepMind acknowledges that objects can drift and interactions last only a few minutes.
In DeepMind’s demonstrations, users move through generated environments with keyboard and mouse controls, then enter instructions to alter the scene, such as changing the weather or adding an object. The model generates subsequent frames in response to those actions rather than rendering a fixed video. DeepMind describes Genie 4 as a world model, not a conventional game engine with a stored 3D map.
DeepMind says Genie 4 can generate ambient sound, music and dialogue together with the visuals. It also describes more natural-looking people and simulations of events such as collisions and aerodynamics. Those capability descriptions come from Google’s own announcement and selected clips; the company has not published benchmark figures in the source material.
DeepMind says Gemini 3.8 Flash can act as a guiding agent, breaking goals into steps and monitoring generated frames. A robotics-related demonstration showed a simulation involving Apptronik’s Apollo humanoid playing checkers. It was a simulation, not a test involving a physical robot. DeepMind says access is planned for trusted testers, with Google AI Ultra subscribers through Project Genie and early Gemini API partners mentioned as future users. No launch date or API price has been specified.
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
Why Walkable AI Worlds Matter
Genie 4’s significance is its attempt to make generated scenes interactive rather than fixed. If users can steer a scene and revise it through ordinary language, the approach could give filmmakers a way to explore previsualization ideas and game teams a tool for early prototypes. DeepMind says testers have reported those uses, but the examples are not evidence of broad commercial adoption or finished production workflows.
The same work is relevant to robotics research because generated environments could help teams explore scenarios without staging every one in the physical world. In the cited Apollo example, however, the robot and task were represented in a simulation. It does not establish that Genie 4 can train a real robot reliably. For viewers, the combination of realistic visuals and matching audio also increases the importance of identifying synthetic media. Google says Genie 4 outputs carry an invisible SynthID watermark.
The practical stakes depend on reliability, access and cost. A world that looks plausible but changes its objects or layout as a user explores it may be useful for brainstorming, yet unsuitable for tasks requiring repeatable physics or stable environments. Those limits are especially relevant to developers deciding whether the model can support work beyond demonstrations.
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 Mechanical Mini Mac Wireless Keyboard Low-Profile – Space Grey
As an affiliate, we earn on qualifying purchases.
From Genie 3 to Genie 4
Genie 4 follows a short series of DeepMind world-model releases. The first Genie, introduced in February 2024, generated simple 2D worlds. Genie 2, presented in December 2024, could create 3D worlds from a single image and maintain them for up to a minute, according to the source material, though real-time operation was limited to a reduced-quality version.
In August 2025, DeepMind described Genie 3 as its first real-time world model, running at 720p and 24 frames per second and maintaining consistency for a few minutes. Google later made Genie 3 available to paying US subscribers through Project Genie in January 2026. The source reports a 60-second limit per world in that service, attributed to compute costs. Genie 4’s announcement presents a further step toward interactive scenes with generated audio, but does not provide public performance benchmarks for comparing the versions.
A key distinction from a standard game engine is that Genie 4 does not rely on a complete, fixed 3D map, according to DeepMind’s description. It predicts what the next moment should look like from earlier frames and user input. That design helps explain both the fluid demonstrations and the company’s warning that details may shift over time.
“Details and objects “can shift or drift over time.””
— Google DeepMind
Open Questions on Reliability and Access
Genie 4 has not been opened to the general public in the announcement described here. DeepMind gives no firm date for Project Genie access or early Gemini API availability, and it has not announced an API price. The company has also not released benchmark results that would let independent readers compare its claims about realism, physics, sound quality or consistency.
How long particular scenes remain stable is not fully settled in the material. DeepMind says interaction can continue for a few minutes; TechCrunch and Bloomberg, as cited by heise, have reported scenes lasting up to about two minutes. That figure is not stated in DeepMind’s own announcement, so it should not be treated as a company-confirmed limit. The demonstrations are selected by DeepMind, and independent testing of the product’s performance is not yet available.
It also remains unclear how reliably the model will handle repeated actions, complex instructions or the same scene across longer sessions. The Apollo example was a research simulation rather than a physical-robot evaluation. The source material does not establish how the SynthID watermark can be detected in every use case or how the announced capabilities will perform outside the company’s demonstrations.
Access and Independent Testing
The next visible milestone is access for trusted testers, followed by the planned availability for Google AI Ultra subscribers through Project Genie and early Gemini API partners. DeepMind has not supplied a release timetable or pricing details for API access. Those announcements should clarify who can test the model and what limits apply outside the company’s demonstrations.
Independent hands-on testing will be needed to establish how stable worlds remain, whether live edits work consistently, and how useful the generated audio and physics are for real projects. Until then, Genie 4 is best described as a demonstrated research and product development direction—not a broadly available tool with independently verified performance.
Key Questions
What is Google Genie 4?
Genie 4 is a Google DeepMind world model that generates an evolving scene in response to user movement and text instructions. Unlike a conventional game engine, DeepMind says it does not use a complete stored 3D map.
Can the public use Genie 4 now?
No public launch is specified. DeepMind says access is planned for trusted testers, with Google AI Ultra subscribers in Project Genie and early Gemini API partners among the intended groups. No date has been announced.
How long can a Genie 4 scene run?
DeepMind says the model supports a few minutes of continuous interaction and warns that details may drift. Reports of scenes lasting about two minutes come from TechCrunch and Bloomberg as cited by heise, not from DeepMind’s own stated limit.
Does Genie 4 involve a real robot?
The Apollo humanoid example described in the source was a simulation. It does not show a physical Apollo robot performing the task or prove that Genie 4 can train real robots reliably.
What remains unproven about Genie 4?
DeepMind has not published benchmark figures in the source material, and the public has not yet had access for independent testing. The model’s reliability, scene stability, practical usefulness and performance beyond selected demonstrations remain unclear.
Source: ThorstenMeyerAI.com
