🔍 Read the full analysis: Google’s Genie 4: Exploring What Its AI Worlds Can And Can’t Do on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google DeepMind presented Genie 4 on Oct. 27, describing an AI system that generates interactive scenes, sound and changes in response to user input. The system is not publicly available; DeepMind says its worlds can drift and support only a few minutes of continuous interaction, and it has released no benchmark results.
Google DeepMind presented Genie 4 on October 27, describing a real-time AI model that generates interactive scenes and sound as users move through a world or give text instructions. The preview marks a new step in the company’s world-model work, but Genie 4 is not publicly available; DeepMind says its generated environments can drift and support only a few minutes of continuous interaction.
Unlike a conventional video generator that renders a clip from a prompt, Genie 4 is designed to predict the next part of a scene in response to controls and instructions. DeepMind says users can move through generated environments with a keyboard or mouse and issue live prompts to change elements such as weather, camera movement or characters. The system generates visuals and audio together, including ambient sounds, music and dialogue, according to the company.
DeepMind also says Gemini models can guide the experience by breaking goals into steps and monitoring the generated frames. Its announcement describes simulations involving events such as collisions and aerodynamics. These are company descriptions and selected demonstrations; DeepMind did not publish benchmark figures in the supplied material, and the demos do not establish how reliably the system performs across different prompts or extended sessions.
Access remains restricted to trusted testers. DeepMind says Genie 4 is expected to come to Google AI Ultra subscribers through Project Genie and to early Gemini API partners, but it has not given a launch date or API price. The source reports that the Ultra subscription costs $200 a month; that is not a Genie 4-specific price. The company says 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
Uses for Prototyping and Simulation
Interactive generated environments could let game developers and filmmakers test ideas without first building every setting by hand. DeepMind says testers have used Genie for film previsualization and game prototypes, but those accounts are feedback from the company’s testing program, not independently measured productivity gains.
The technology also has potential relevance to robotics research, where simulated environments can provide settings for testing tasks. DeepMind showed a demonstration involving a simulated Apptronik Apollo humanoid playing checkers. No physical robot was involved in that demonstration, according to the source material, so it should not be read as evidence that Genie 4 has trained or validated a real robot. Waymo, meanwhile, already uses the earlier Genie 3 in its driving simulator, according to the source.
For viewers, increasingly convincing generated scenes raise questions about how synthetic media is identified. DeepMind says SynthID is embedded in Genie 4’s audio and video output. The company has not provided details here on how the watermark can be detected or how well it persists after editing or reposting.
Products Worth Considering
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Samsung 32-Inch Odyssey OLED G8 (G80SH) 4K Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Genie Builds on Earlier Models
Genie is part of DeepMind’s effort to build models that generate environments over time rather than produce only a finished image or video. The initial Genie research, published in February 2024, described an 11-billion-parameter model trained on unlabeled internet videos to create simple two-dimensional worlds. Genie 2, announced in December 2024, generated 3D environments from a single image and could keep them consistent for up to a minute, while real-time use required a reduced 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 some paying U.S. subscribers through Project Genie, according to the supplied account. That history gives context for Genie 4’s proposed improvements, but does not independently verify the new model’s claimed gains in realism, sound or control.
A key distinction is that Genie generates what comes next rather than relying on a stored, complete 3D map, as a conventional game engine does. DeepMind says the model draws on what it has already generated to predict later frames. That approach can create responsive scenes, but the company also acknowledges that objects and details may shift or drift during longer interactions.
“Details and objects “can shift or drift over time.””
— Google DeepMind, describing Genie 4’s limits
Open Questions on Access and Reliability
Several practical details remain undisclosed. DeepMind has not announced a public release date, API pricing, benchmark results or independent evaluations in the source material. The company has also not defined exactly how long a typical session lasts or how performance changes across different scenes, controls and prompts.
Some reports cited in the source put demonstrated scenes at about two minutes, but that duration is not stated in DeepMind’s own supplied description and should not be treated as an official limit. It is also unclear how often users will encounter visual drift, whether generated worlds preserve important details after revisiting a location, and how accessible the planned service will be beyond trusted testers and eligible subscribers.
The demos are selected by DeepMind, so they show what the company chose to present rather than a representative range of results. Claims about improved physics, natural people and practical applications have not been accompanied here by third-party testing. SynthID’s presence is a company claim; the supplied material does not explain its detection process or effectiveness after content has been altered.
Release Plans and Testing Ahead
DeepMind says it plans to bring Genie 4 to Google AI Ultra subscribers through Project Genie and to early Gemini API partners. No timetable or API price has been announced, leaving the next public milestone unclear. Broader access would allow users and independent reviewers to test whether the model’s live controls, audio generation and scene consistency hold up outside selected demonstrations.
Until then, the main evidence is DeepMind’s presentation and the limited tester feedback it has described. Any assessment of commercial use, including game development, film planning or robotics research, will depend on access, costs, reliability and results from testing beyond the company’s own examples.
Key Questions
What is Google Genie 4?
Genie 4 is a world model that DeepMind says generates scenes over time in response to user movement and instructions. It is intended to create interactive environments rather than only render a fixed video clip.
Can the public use Genie 4 now?
No. The source says access is limited to trusted testers. DeepMind plans access through Project Genie for Google AI Ultra subscribers and for early Gemini API partners, but has not announced a date.
What can users do inside a Genie 4 world?
DeepMind says users can navigate with keyboard or mouse controls and use text prompts to change scene elements, including weather or objects. The model can also generate audio alongside the visuals, according to the company.
What are Genie 4’s known limitations?
DeepMind says details and objects may shift or drift over time and describes interaction as lasting a few minutes. The company has not published benchmark results or a more precise general limit in the supplied material.
Does the robotics demonstration show Genie 4 controlling a real robot?
No. The cited demonstration showed a simulated Apptronik Apollo humanoid playing checkers; the source says no real robot was involved. It is a research demonstration, not evidence of real-world robot performance.
Source: ThorstenMeyerAI.com
