🔍 Read the full analysis: Can Genie 4 Build Worlds You Can Actually Explore? 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, describing a real-time model that generates interactive scenes, sound and changes prompted by text. The demonstrations are company-selected, access is limited to trusted testers, and DeepMind says worlds can drift after a few minutes.
Google DeepMind demonstrated Genie 4 on October 27, a real-time AI model that generates scenes users can move through and change with text prompts. The company says it also creates synchronized audio, but the system is available only to trusted testers, and its demonstrations do not establish how reliably it works beyond short sessions.
DeepMind describes Genie 4 as a world model: rather than rendering a scene from a stored 3D map like a conventional game engine, it predicts what should appear next based on the scene so far and a user’s inputs. Demonstrations show keyboard and mouse controls alongside instructions to change elements such as weather or add objects. These capabilities are presented in DeepMind’s own selected clips, not independent tests.
The company says Genie 4 generates sound alongside images, including ambient noise, music and dialogue, and produces more realistic people and physical interactions. It also describes Gemini 3.8 Flash as a guiding system that interprets goals, breaks them into steps and monitors the generated scene. DeepMind has not published benchmark results in the material provided, so the performance claims cannot be compared quantitatively with other systems.
DeepMind says the model can support a few minutes of continuous interaction, while warning that details and objects may shift or drift over time. TechCrunch and Bloomberg, as cited by heise, have reported scenes lasting about two minutes; that specific duration is not in DeepMind’s stated description. Access is planned for Google AI Ultra subscribers through Project Genie and early Gemini API partners, but the company has given no launch date or API price.
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 dependable tools, Genie 4 could let teams test ideas in responsive virtual environments without first building every scene in a conventional game engine. DeepMind says testers have used the model for film previsualization and game prototypes. Those examples are reported tester feedback, not evidence of established production use or broad availability.
The potential extends to simulation, but the evidence needs careful distinction. DeepMind showed a robotics research demonstration in which a simulated Apptronik Apollo humanoid played checkers; no physical robot took part, according to the source material. Waymo already uses the predecessor, Genie 3, in its driving simulator, giving a concrete example of a related model’s research or operational relevance without proving Genie 4 is ready for that role.
Generated scenes that include matching sound may also make synthetic video more convincing to viewers. DeepMind says Genie 4 video and audio outputs carry an invisible SynthID watermark. The source does not provide independent verification of the watermark’s coverage or how readily it can be detected in every setting.
Popular right now
Samsung 42-Inch Class OLED S90F 4K Smart TV (2025)
Dyson V8 Cyclone Cordless Vacuum
Shark Navigator Lift-Away Upright Vacuum (NV360)
As an affiliate, we earn on qualifying purchases.
How Genie Reached Real Time
Genie 4 follows a series of DeepMind systems that have moved from simple generated environments toward real-time interaction. The first Genie, introduced in February 2024, generated simple 2D worlds; its research paper described an 11-billion-parameter model trained on unlabeled internet videos. Genie 2, presented in December 2024, generated 3D worlds from a single image and maintained consistency for up to a minute, while real-time use required a reduced-quality version.
In August 2025, Genie 3 became DeepMind’s first world model to run in real time. The company described output at 720p and 24 frames per second, with consistency for a few minutes and visual memory reaching back about one minute. Google opened Genie 3 to paying U.S. subscribers through Project Genie in January 2026. TechCrunch reported a 60-second limit per world, attributed partly to computing costs. Genie 4’s announced plans build on that consumer-facing project, but the provided material does not confirm that the earlier access limit or pricing will carry over.
“Details and objects “can shift or drift over time.””
— Google DeepMind
Limits Behind the Demonstrations
Important details remain unsettled because Genie 4 is not generally available and the demonstrations were selected by DeepMind. The company has not supplied independent evaluations, benchmark figures or enough public access to establish how consistently users can navigate a scene, preserve objects and apply live instructions over longer sessions.
The exact session limit is also unclear. DeepMind says “a few minutes,” while outside reports cited in the source describe scenes of about two minutes. The duration may vary by setup, but the available material does not explain how it is measured or what happens when a session reaches its limit. No date has been announced for the planned subscriber or API access, and the API price and availability terms have not been disclosed.
Use cases should be treated as early-stage evidence. Film and game examples are tester feedback reported by DeepMind; the Apollo checkers scene was a simulation, not a physical-robot trial. The source also does not establish whether SynthID remains detectable after editing, compression or other transformations.
Access Plans and Testing Ahead
The next milestone is broader access. DeepMind says Genie 4 is expected to reach Google AI Ultra subscribers through Project Genie and early Gemini API partners, but has not announced a date or price for API use. Until access expands, outside developers and researchers cannot independently assess the model through the planned channels.
Further public details would help establish how long worlds can run, how stable they remain, what computing resources they require and how the sound and watermarking perform in practice. For now, the confirmed development is a company demonstration and a limited testing plan—not a generally available world-building tool.
Key Questions
What is Genie 4?
Genie 4 is a world model that DeepMind says generates scenes frame by frame in response to user movement and instructions. Unlike a conventional game engine, it does not use a stored 3D map, according to the company’s description.
Can people try Genie 4 now?
Not generally. DeepMind says access is limited to trusted testers, with planned access for Google AI Ultra subscribers through Project Genie and early Gemini API partners. No public release date has been announced.
How long can a Genie 4 world run?
DeepMind says Genie 4 supports a few minutes of continuous interaction and warns that objects can drift. Outside reports cited in the source put scenes at about two minutes, but DeepMind’s description does not give that exact figure.
Does Genie 4 generate audio?
DeepMind says the model generates audio with its visuals, including ambient sounds, music and dialogue. The claim has been shown in the company’s demonstrations; the provided material does not include independent testing.
Is the robotics demonstration a real robot test?
No. The Apollo checkers example was a simulation, and the source says no physical robot was involved. It should not be read as evidence that Genie 4 has been tested controlling a real Apollo robot.
Source: ThorstenMeyerAI.com
