🔍 Read the full analysis: A Guide To Genie 4’S World-Building AI And Who Can Access It 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 introduced Genie 4 on October 27, 2026, describing it as a real-time AI model that generates interactive scenes, including audio, in response to movement and text prompts. It is not publicly available: Google says access is planned for trusted testers, with no release date or API pricing announced. The company also says generated scenes can drift over time and support only a few minutes of continuous interaction.
Google DeepMind demonstrated Genie 4 on October 27, describing it as a real-time AI model that builds interactive worlds from user input, including typed instructions, movement controls and generated sound. The system is not publicly available: the company says it is being tested by trusted users and plans to offer access to Google AI Ultra subscribers through Project Genie and to early Gemini API partners, but has not announced dates or API prices.
DeepMind says Genie 4 generates each new moment of a scene as a user moves through it or types a change, rather than drawing from a stored 3D map or conventional game engine. In its example demonstrations, users steer scenes with keyboard and mouse controls and request changes such as rain, a camera movement, or a new person or object. The model then generates subsequent frames in response.
Audio is part of the generated scene, according to DeepMind. The company says the model can produce sounds such as wind and water, music and character dialogue alongside its visuals. DeepMind also says Gemini 3.8 Flash can interpret user goals, break them into steps and monitor the resulting scene. These capabilities are the company’s descriptions of its system; it has not published benchmark figures in the source material.
DeepMind describes Genie 4’s limits as well. It says details and objects can shift or drift over time and that the model supports “a few minutes of continuous interaction.” TechCrunch, and Bloomberg as cited by heise, reported clips lasting about two minutes; that duration is not specified in DeepMind’s own text. The demonstrations were selected and edited by the company, so they do not establish how the system performs across unrestricted use.
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
Genie 4 points toward AI systems that respond continuously to a user, rather than producing a finished video from a single prompt. If the approach works reliably, creators could use generated scenes to explore visual ideas or test game concepts without first building a conventional 3D environment. For now, those applications remain limited: DeepMind describes film previsualization and game prototyping as feedback from testers, not as independently verified production results.
The system also has potential research uses, but the demonstrations should not be mistaken for completed real-world testing. DeepMind showed a robotics experiment in which a model broke a task into steps and Genie 4 simulated Apptronik’s Apollo humanoid playing checkers. The Decoder reported that no physical robot was involved in that demonstration. Separately, Waymo uses the earlier Genie 3 model in its driving simulator, according to the source material; that does not establish that Waymo is using Genie 4.
Generated sound and increasingly realistic imagery make the model’s provenance relevant to viewers and publishers. Google says Genie 4 video and audio outputs carry an invisible SynthID watermark. The source does not explain how readily that watermark can be detected after editing or how the system performs outside the company’s chosen examples.
Products Worth Considering
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
LG UltraGear 32GX870A-B 32-inch 4K OLED Gaming Monitor
LG UltraGear 27G640A-B 27-Inch QHD IPS Gaming Monitor, 300Hz, 1ms, G-SYNC Compatible, FreeSync Premium, DisplayHDR 400
As an affiliate, we earn on qualifying purchases.
How Genie Models Have Evolved
Genie 4 follows a series of DeepMind world-model releases. 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, presented in December 2024, generated 3D worlds from a single image and maintained them for up to a minute, though real-time operation required a reduced-quality version.
In August 2025, DeepMind presented Genie 3 as its first world model to run in real time, at 720p and 24 frames per second, with scenes remaining consistent for a few minutes and visual memory extending about a minute. Google opened Genie 3 through Project Genie to paying U.S. subscribers in January 2026. TechCrunch reported a 60-second limit per world at that time, partly because of computing costs. Genie 4’s announced audio generation and live text changes are presented by DeepMind as new capabilities, but the company has not provided comparative benchmark results in the supplied material.
“details and objects can shift or drift over time”
— Google DeepMind
Access and Performance Still Unclear
No public launch date has been announced, and the source gives no price for Gemini API access. It is also unclear how many testers currently have access, what eligibility requirements will apply, or whether Project Genie access will be limited by geography, subscription tier or usage caps. The source identifies Google AI Ultra, described there as a $200-a-month tier, as the planned route for subscribers; Google has not supplied a specific timetable.
Performance beyond the selected demonstrations remains uncertain. DeepMind has not published benchmark results, and its stated interaction limit and warning about drift leave open how well the system handles longer sessions or complex user instructions. The reported clips of about two minutes come from media coverage, not the company’s own stated duration. The watermark’s practical detection and durability also have not been detailed in the supplied material.
Awaiting Tester Rollout Details
The next expected step is access for trusted testers, followed by the planned availability for Google AI Ultra subscribers through Project Genie and early partners using the Gemini API. Google has not set a date for either route or disclosed API pricing. Further details on eligibility, session limits, geographic availability and product terms will be needed before prospective users can know how broadly they can use the model.
As access expands, independent testing will help establish whether Genie 4’s generated worlds remain coherent beyond short, curated demonstrations and how useful its audio and text controls are for creative or research work. Until then, the company’s clips and statements describe the announced capabilities, not independently measured performance.
Key Questions
What is Genie 4?
Genie 4 is a world model that Google DeepMind says generates interactive scenes as users move through them or issue typed instructions. The company says it can generate accompanying audio, too.
Can the public use Genie 4 now?
No. DeepMind says the model is currently being tested by trusted users. Planned access includes Google AI Ultra subscribers through Project Genie and early Gemini API partners, but no launch date has been announced.
How long can a Genie 4 scene run?
DeepMind says the model supports a few minutes of continuous interaction and warns that objects or details can drift. Some media reports describe scenes of about two minutes, but that figure is not in the company’s own stated description.
Can Genie 4 train a real robot?
The cited Apollo demonstration was a simulation, not a test involving a physical robot, according to The Decoder. The source material says Waymo uses the earlier Genie 3 model in a driving simulator, but does not say Waymo uses Genie 4.
Does Genie 4 add watermarks to its videos?
Google says every Genie 4 video and audio output carries an invisible SynthID watermark. The supplied material does not explain how the watermark can be detected or whether it remains intact after editing.
Source: ThorstenMeyerAI.com
