🔍 Read the full analysis: How To Think About Google’s Genie 4: Worlds, Weak Spots, And Access on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
Google DeepMind presented Genie 4 on Oct. 27, describing a real-time AI model that generates interactive scenes with sound and responds to typed instructions. The company has not released it publicly, published benchmark results or set an access date; it says scenes can drift and continuous interaction lasts a few minutes.
Google DeepMind presented Genie 4 on Oct. 27, describing it as a real-time AI “world model” that generates scenes users can move through and alter with text instructions. The announcement points to possible uses in game prototyping, film previsualization and robotics research, but the system is not generally available, and DeepMind says scenes can drift during sessions that last only a few minutes.
In DeepMind’s demonstrations, users navigate generated scenes with keyboard and mouse controls and enter prompts to change conditions, such as making it rain. The model generates images and audio together, including ambient sounds, music and dialogue, according to the company. DeepMind also says a Gemini model helps interpret instructions, break goals into steps and monitor the generated scene.
Those capabilities are DeepMind’s descriptions of its own system, shown in company-selected demonstrations. The announcement materials cited by the source report no benchmark results. The demonstrations therefore show what the company chose to present, not an independent evaluation of performance across tasks or users.
DeepMind says access is currently limited to trusted testers. It plans to offer Genie 4 to Google AI Ultra subscribers through Project Genie and to early Gemini API partners, but has not provided an availability date or API pricing. The source identifies Google AI Ultra as a $200-per-month subscription tier at the time of publication.
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
Potential Uses and Practical Limits
Genie 4 could let developers test environments by describing changes rather than building every setting with conventional game tools. DeepMind has reported tester feedback about game prototypes and film previsualization. That feedback is not evidence that the system is ready for production or that it is faster or cheaper than existing tools.
The robotics example also needs qualification. DeepMind describes a simulated task involving a humanoid robot, but the source says no physical robot took part. Synthetic environments may be useful for research, yet a simulation does not establish how well a robot will perform in the real world. The source says Waymo uses Genie 3, a predecessor, in its driving simulator; that does not establish that Waymo is using Genie 4.
For viewers, realistic scenes with generated sound may make synthetic media harder to recognize by sight and hearing alone. DeepMind says Genie 4 outputs carry an invisible SynthID watermark. How reliably that watermark can be detected or retained after content is edited is not established in the supplied material.
Popular right now
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Samsung 42-Inch S90H OLED 4K Smart TV Bundle
Vitamix Propel Series 510 Professional-Grade Blender, 3 Programs, Black
As an affiliate, we earn on qualifying purchases.
How Genie Has Developed
Genie 4 follows earlier systems in DeepMind’s Genie series. The source says the first Genie, described in a February 2024 research paper, generated simple 2D worlds. Genie 2, announced in December 2024, generated 3D environments from a single image and maintained them for up to a minute, though real-time operation was limited to a lower-quality version.
Genie 3, announced in August 2025, was DeepMind’s first real-time world model, according to the source. The company reported 720p video at 24 frames per second and consistency for a few minutes, with visual memory reaching back about a minute. Project Genie later gave paying U.S. subscribers access to Genie 3; TechCrunch reported a 60-second limit per world, citing compute costs. These predecessor figures provide context, but they are not independent measurements of Genie 4.
A world model differs from a conventional video generator because it produces successive frames in response to user actions, rather than rendering a fixed clip. It also differs from a game engine: the source says Genie 4 does not rely on a stored 3D map of the scene, but predicts what comes next from what it has generated and the user’s input. That approach can support flexible scenes, while leaving room for objects and details to change as the session continues.
“Details and objects “can shift or drift over time.””
— Google DeepMind, describing Genie 4’s limits
Benchmarks and Access Remain Open
Key details remain unavailable. DeepMind has not announced benchmark results, a public launch date, API pricing or a specific release timetable. The exact session length and the conditions under which a scene begins to drift are also not specified in the supplied material.
Reports cited in the source from TechCrunch and Bloomberg, via heise, describe Genie 4 scenes lasting up to about two minutes. That duration is not stated in DeepMind’s own text and should not be treated as a confirmed company specification. The source also does not provide independent testing of the model’s physics, audio quality or ability to keep scenes consistent.
It is not yet clear how broad the planned tester access will be, which Gemini API partners will participate, or whether the demonstrations’ capabilities will be available consistently in real use. Claims about film, games and robotics remain early indicators rather than verified commercial outcomes.
Awaiting Tester and Pricing Details
The next stated step is access for trusted testers, Google AI Ultra subscribers through Project Genie and early Gemini API partners. DeepMind has not given a date for those wider access routes, so readers will need to wait for an availability announcement and details of any product limits.
Further reporting should look for information on session duration, costs, benchmarks and how well the model maintains a scene across repeated interactions. Independent access would help distinguish the abilities shown in selected demonstrations from performance users can expect in practical work. Until then, Genie 4 is best understood as a company-announced system with promising demonstrations and acknowledged limits, not a broadly available product.
Key Questions
What is Google Genie 4?
Genie 4 is an AI world model that DeepMind says generates interactive visual scenes in response to user movement and text instructions, with audio generated alongside the scene.
Can the public use Genie 4 now?
No public release is described in the supplied material. DeepMind says it is available to trusted testers and plans access through Project Genie for Google AI Ultra subscribers and for early Gemini API partners, without giving a date.
How long can a Genie 4 scene run?
DeepMind says Genie 4 supports a few minutes of continuous interaction. Reports of scenes lasting up to about two minutes are attributed to outside coverage in the source, not to DeepMind’s own stated specification.
What are Genie 4’s known limitations?
DeepMind says details and objects may shift or drift over time, and it describes sessions as lasting a few minutes. The supplied announcement details include no benchmark results for scene consistency or other capabilities.
Does Genie 4 train or control real robots?
The source describes a robotics simulation, not a test involving a physical robot. DeepMind’s example shows a simulated humanoid task and does not establish real-world robot performance.
Source: ThorstenMeyerAI.com
