🔍 Read the full analysis: How To Explore Google’s Genie 4 And Understand Its Current Limits 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 introduced Genie 4 on October 27, 2026, describing it as a real-time AI model that creates interactive scenes with 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 announced Genie 4 on October 27, describing it as a real-time world model that generates scenes people can steer with keyboard, mouse and text. The demonstrations show visuals and audio generated together, but the system is not yet broadly available, and DeepMind says scenes may drift during sessions that last only a few minutes.
DeepMind says users can move through a generated scene and enter instructions to change it, such as altering the weather or adding an object. The model predicts what happens next from the scene so far and the user’s inputs, rather than rendering a world from a stored 3D map like a conventional game engine. The company says its demonstrations include ambient sound, music and dialogue, as well as scenarios involving people and physical interactions.
DeepMind also describes Gemini 3.8 Flash as a guiding agent that watches the scene, breaks goals into steps and helps steer the generated world. The company says the system can simulate events such as collisions and aerodynamics. Those capabilities are vendor descriptions, not independently verified results: DeepMind has not published benchmark figures, and the public examples were selected by the company.
Access is currently limited to trusted testers. DeepMind says Genie 4 is coming to Google AI Ultra subscribers through Project Genie and to early Gemini API partners, but it has not announced an availability date or API price. The source report identifies Google AI Ultra as a $200-per-month subscription tier. Reported tester applications include film previsualization and game prototypes; the robotics example is a research demonstration, not evidence of a robot trained or operated in the real world.
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 Interactive Worlds Matter
If the demonstrations translate into reliable tools, Genie 4 could let creators and developers test ideas in scenes that respond to directions rather than requiring a finished video or a fully built game environment. DeepMind-reported feedback points to film previsualization and game prototyping as possible uses. For researchers, simulated environments may help explore tasks and scenarios before testing them with physical equipment.
That potential needs to be weighed against the limits in the announcement. A scene that changes unpredictably or loses continuity may be unsuitable for precise production work, and no published benchmarks currently show how often the model follows instructions or maintains a scene. The generated audio and lifelike imagery also make provenance relevant: DeepMind says Genie 4 video and audio carry an invisible SynthID watermark. That is the company’s stated safeguard, not an independent assessment of how reliably outputs can be identified.
Popular right now
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
eufy E25 Robot Vacuum and Mop Combo with HydroJet Roller Mop
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
As an affiliate, we earn on qualifying purchases.
From Genie 3 to Genie 4
Genie 4 follows a series of DeepMind world-model projects. The first Genie, announced in February 2024, generated simple 2D worlds; its research paper described an 11-billion-parameter model trained on unlabeled internet videos. Genie 2, announced in December 2024, created 3D worlds from a single image and maintained them for up to a minute, though its real-time version was lower quality.
In August 2025, DeepMind presented 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 opened Genie 3 through Project Genie to paying US subscribers, according to the source report, which cited a 60-second session limit reported by TechCrunch. Genie 4’s described addition of native audio and live text-directed changes marks a further step, but direct performance comparisons are difficult without published common tests.
DeepMind’s robotics demonstration depicts a simulated Apptronik Apollo humanoid playing checkers. As reported by The Decoder, no physical robot took part. Waymo’s use of the predecessor Genie 3 in its driving simulator offers a separate example of how Google has applied earlier world-model research, but it does not establish Genie 4’s readiness for commercial or safety-critical use.
“Details and objects “can shift or drift over time.””
— Google DeepMind
Access, Duration and Reliability
Several practical details remain unknown. DeepMind has not given a public release date, specified the Gemini API price or published benchmark results for realism, instruction-following, scene continuity or audio quality. It is also unclear how much access testers will receive and whether they will be able to create, save or share worlds beyond the selected demonstrations.
DeepMind describes interaction as lasting “a few minutes,” while TechCrunch and Bloomberg, as cited by heise, reported scenes lasting up to about two minutes. That figure is not in DeepMind’s own announcement, and the difference may reflect separate reporting or test conditions. The source material does not establish a consistent public session limit. Claims about potential uses should likewise be treated cautiously: reported creator feedback is not the same as demonstrated production performance, and a simulated robot task is not a physical-world trial.
When Wider Testing Begins
The next milestone is wider access for Google AI Ultra subscribers through Project Genie and early partners using the Gemini API. DeepMind has said those openings are coming soon but has not provided dates or commercial terms. Until access expands, most people can assess Genie 4 only through DeepMind’s selected demonstrations, not by independently testing its controls or limitations.
Further evidence will depend on hands-on evaluations and technical details from the company or testers. Useful questions include whether scenes remain consistent across longer sessions, how accurately text instructions change them, and how sound and visual output behave under repeated use. For now, Genie 4 is an announced research and development system with striking demos and acknowledged limits, rather than a generally available tool whose performance can be independently checked.
Key Questions
Can the public try Genie 4 now?
No. DeepMind says access is currently limited to trusted testers. It has described plans for Google AI Ultra subscribers through Project Genie and early Gemini API partners but has not announced dates.
What does Genie 4 do?
DeepMind describes Genie 4 as a world model that generates a scene as a user moves through it and gives instructions. The company says it can produce visuals and audio together and respond to live text prompts.
How long can a Genie 4 session last?
DeepMind says the model supports “a few minutes” of continuous interaction. TechCrunch and Bloomberg, as cited by heise, reported scenes of up to about two minutes, but that specific limit is not in DeepMind’s own announcement.
Does Genie 4 train or operate real robots?
The example described in the source is a robotics simulation involving an Apptronik Apollo humanoid playing checkers; it did not involve a physical robot. The demonstration does not establish that Genie 4 has trained or controlled a real robot.
How can viewers identify Genie 4 videos or audio?
DeepMind says Genie 4 video and audio outputs carry an invisible SynthID watermark. The announcement does not provide independent evidence about how reliably the watermark can identify all outputs.
Source: ThorstenMeyerAI.com
