🔍 Read the full analysis: Who Can Walk Into A World Built By Google’s Genie 4? on ThorstenMeyerAI.com
Get the latest gadgets 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 it as a real-time model that generates interactive scenes with sound and responds to user controls and text prompts. It is not publicly available: access is limited to trusted testers, with broader access promised for Google AI Ultra subscribers and Gemini API partners but no date announced. DeepMind says sessions last a few minutes and scenes can drift; the demos are company-selected, with no published benchmark results.
Google DeepMind demonstrated Genie 4 on October 27, describing it as a model that generates interactive scenes in real time and can add sound or respond to typed instructions. The system is not yet open to the public: the company says it is available to trusted testers, with access planned for Google AI Ultra subscribers through Project Genie and early Gemini API partners, but it has not given a launch date.
In DeepMind’s demonstrations, users move through generated scenes with keyboard and mouse controls and can type changes while a scene runs, such as changing the weather or adding an object. The company says Genie 4 creates video and audio together, including environmental sounds, music and character dialogue. These capabilities are presented in selected company demos; DeepMind has not published benchmark figures for them.
DeepMind describes the system as a world model: it generates what happens next from the scene so far and the user’s input, rather than rendering a view from a stored 3D map like a conventional game engine. The company says Gemini 3.8 Flash helps interpret goals, break them into steps and monitor the generated scene. The source material also describes a robotics demo involving a simulated Apptronik Apollo robot playing checkers; no physical robot took part.
The company’s own stated limitations include scenes that can shift or drift over time and interaction lasting “a few minutes.” TechCrunch and Bloomberg, as cited by heise, have reported scenes of about two minutes, but that specific duration does not appear in DeepMind’s announcement, according to the source material. Google says Genie 4’s 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
Access and Uses Remain Limited
Genie 4 points toward AI systems that respond to a person’s actions instead of producing only a finished clip. If the demonstrated controls and sound work reliably beyond selected examples, such systems could give game developers and filmmakers a way to sketch interactive settings or test ideas without building a complete game environment. DeepMind has cited tester feedback about film previsualization and game prototypes, but those examples are not evidence of broad commercial deployment.
The distinction between a simulated world and a physical one also matters. The checkers example is a research demonstration, not proof that a robot can safely learn or perform the task in the real world. At the same time, increasingly realistic generated footage and audio may complicate judgments about whether a clip is authentic. Google’s stated use of SynthID is relevant to that concern, although the source material does not provide independent testing of how the watermark performs.
For readers, the immediate practical point is access: despite the public demonstrations, Genie 4 is not currently a product people can try freely. Its cost, availability dates and technical performance outside curated examples remain unknown.
Products Worth Considering
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
As an affiliate, we earn on qualifying purchases.
From Genie 3 to Interactive Worlds
The announcement builds on earlier versions of DeepMind’s Genie models. The first Genie, introduced in February 2024, generated simple 2D worlds from video training data. Genie 2 followed in December 2024, generating 3D worlds from a single image and maintaining consistency for up to a minute, according to the source material. Its real-time operation required a lower-quality version.
In August 2025, DeepMind introduced Genie 3 as its first real-time world model, reporting 720p output at 24 frames per second and consistency 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 for that service, citing compute costs. Those Genie 3 details do not establish Genie 4’s limits; the new model’s access terms and performance have not been fully specified.
DeepMind’s description of Genie 4’s architecture is also different from a conventional game engine. It predicts future frames from previous ones rather than consulting a fixed, complete 3D map. That approach can support responsive scenes, but it also helps explain why visual details may drift as an interaction continues.
“Details and objects “can shift or drift over time.””
— Google DeepMind, describing Genie 4’s limits
Duration, Quality and Release Date
Several practical details remain unsettled. DeepMind has not announced when Genie 4 will become available, how much Gemini API access will cost, or what limits will apply to subscribers. “A few minutes” is the company’s description of continuous interaction; the approximately two-minute figure reported by some outlets is not stated in DeepMind’s own text cited here.
The demonstrations also do not establish how often scenes remain coherent, how well text instructions work across varied settings, or how the model performs outside company-selected examples. DeepMind has published no benchmark numbers in the supplied material. It is also unclear whether SynthID can be reliably detected after generated media is edited or redistributed.
Broader Testing Is the Next Milestone
The next stated step is access for trusted testers, followed by planned availability for Google AI Ultra subscribers through Project Genie and early partners using the Gemini API. DeepMind has not supplied a timetable, subscription eligibility details for Genie 4, or API pricing. Those announcements will clarify who can use the model and under what limits.
Further testing will also show whether the interactive scenes remain coherent for the promised few minutes, how useful the audio and text controls are in routine work, and whether the robotics demonstrations translate into research beyond simulation. Until those details are available, Genie 4’s capabilities should be understood as company-described demonstrations, not independently verified performance or a generally available tool.
Key Questions
What is Google DeepMind’s Genie 4?
It is a world model that generates successive frames of a scene in response to user actions and instructions. DeepMind says it can also generate audio alongside the visuals.
Can the public use Genie 4 now?
No. DeepMind says access is currently limited to trusted testers. Broader access is planned for Google AI Ultra subscribers through Project Genie and early Gemini API partners, with no date announced.
How long can a Genie 4 world run?
DeepMind describes continuous interaction as lasting a few minutes and says objects and details can drift. Reports of scenes lasting about two minutes are attributed to other outlets, not the company’s own wording cited here.
Does the robotics demo show a real robot using Genie 4?
No. The cited checkers demonstration simulated an Apptronik Apollo humanoid; the source material says no physical robot was involved.
Has DeepMind independently verified Genie 4’s performance?
The supplied material describes company-selected demonstrations and reports that DeepMind published no benchmark numbers. Its capabilities should therefore be treated as the company’s claims pending broader testing.
Source: ThorstenMeyerAI.com
