🔍 Read the full analysis: Google’s Genie 4 Makes AI Worlds You Can Explore—With Some 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, describing a real-time AI system that generates interactive scenes, sound and changes prompted by text. The system is not generally available; DeepMind says access is planned for selected testers and warns that scenes can drift during sessions lasting a few minutes.
Google DeepMind demonstrated Genie 4 on October 27, a system that generates interactive scenes in response to keyboard or mouse input and text prompts, with audio generated alongside the visuals. The model is not yet generally available: DeepMind says it is being tested by selected users and acknowledges that scenes can shift or drift over time and support only a few minutes of continuous interaction.
In DeepMind’s demonstrations, users move through generated scenes with standard controls and can type instructions to alter them while they run, such as changing the weather or adding an object. The company describes Genie 4 as a world model: rather than rendering a completed clip, it predicts what the next moment should look like based on the scene so far and the user’s actions. It does not rely on a stored, complete 3D map like a conventional game engine.
DeepMind says the system generates sound in real time, including environmental audio, music and dialogue, and models interactions such as collisions and movement. It also describes Gemini 3.8 Flash as an agent that can interpret goals, break them into steps and monitor the frames as the scene changes. Those capabilities were presented in company-selected demonstrations; DeepMind did not publish benchmark figures to independently measure performance.
Access remains limited. DeepMind says Genie 4 is available to trusted testers, with plans to bring it to Google AI Ultra subscribers through Project Genie and to early Gemini API partners. The company has not provided a launch date or API pricing. Reports cited in the source material describe scenes of up to about two minutes, but that figure is not in DeepMind’s own announcement; its stated limit is a few minutes of continuous interaction.
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 Generated Clips to Interactive Scenes
Genie 4 points toward AI systems that generate not just a video, but a scene people can move through and influence. If the approach becomes reliable and less costly to run, it could give game designers and filmmakers a faster way to sketch environments or test ideas. DeepMind reports that testers have used it for film previsualization and game prototypes, but those accounts are feedback from a limited testing group, not evidence of broad commercial adoption.
The distinction matters for robotics too, but the demonstrations should not be mistaken for real-world robot training results. DeepMind showed a simulated task involving an Apptronik Apollo humanoid playing checkers; according to the source material, no physical robot took part. The system may help researchers explore simulated scenarios, but the demonstration alone does not establish that a robot learned or performed the task in the real world.
More convincing generated video and matching audio also raise questions about identifying synthetic media. Google says Genie 4’s video and audio outputs carry an invisible SynthID watermark. That is the company’s stated measure; the announcement does not provide independent testing of how reliably the watermark can be detected after media is edited or shared.
Products Worth Considering
DJI Mic 3 Wireless Microphone System (2 Transmitters, 1 Receiver, Charging Case)
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Genie Reached Version Four
The Genie project has developed from generating simple scenes toward interactive, real-time worlds. The first Genie, introduced in February 2024, created simple 2D environments from video. Genie 2 followed in December 2024, generating 3D worlds from a single image and maintaining consistency for up to a minute, though real-time use required a lower-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 keeping scenes consistent for a few minutes. Google later opened Genie 3 through Project Genie to paying US subscribers in January 2026. TechCrunch reported a 60-second limit per world at that time, citing compute costs. Genie 4’s new emphasis, based on DeepMind’s presentation, is more realistic scenes, integrated audio and text-based changes while interacting.
“Details and objects “can shift or drift over time.””
— Google DeepMind
Access, Reliability and Costs Remain Open
DeepMind has not announced when Project Genie users or Gemini API partners will receive access, how many testers are involved, or what API use will cost. It also has not published benchmark results for visual consistency, audio quality, response speed or the model’s handling of complex instructions. The public demonstrations were selected by the company, so they do not show how often the system fails across ordinary use.
How long a scene stays coherent in practice is also unsettled. DeepMind’s description says a few minutes, while outside reports cited in the supplied material put some scenes at about two minutes. The source does not establish whether those figures reflect a fixed product limit, a test result or different conditions. It remains unclear how much compute each session requires and whether performance or access will vary by plan.
Watch for a Wider Access Date
The next concrete milestone is Google’s planned expansion from trusted testers to Project Genie subscribers and early Gemini API partners. DeepMind has not set a date or released pricing, so interested users and developers will need to wait for those details. Further information about access limits, system performance and safeguards may emerge as testing expands.
For now, the demonstrations establish what DeepMind says Genie 4 can do, not how consistently it performs outside curated examples. Any assessment of its value for production workflows or robotics will depend on broader testing, clearer measures of reliability and evidence that generated scenes remain coherent for the tasks users want to complete.
Key Questions
What is Google Genie 4?
Genie 4 is a world model that predicts and generates a scene as a user moves through it or gives instructions. Unlike a finished video, the scene changes in response to ongoing input.
Can the public use Genie 4 now?
No. DeepMind says it is available to trusted testers and plans access through Project Genie for Google AI Ultra subscribers and through early Gemini API partners. No public release date has been announced.
How long can a Genie 4 scene run?
DeepMind says the system supports a few minutes of continuous interaction and warns that details or objects can drift. Some outside reports cited in the source material describe scenes of about two minutes, but that figure is not stated in DeepMind’s announcement.
Does the robotics demonstration show a real robot using Genie 4?
No. The cited demonstration showed a simulated Apptronik Apollo performing a checkers task. The source material says no physical robot was involved.
How does Google say users can identify Genie 4 media?
Google says video and audio generated by Genie 4 carry an invisible SynthID watermark. The announcement does not provide independent results on how well that watermark survives editing or redistribution.
Source: ThorstenMeyerAI.com
