🔍 Read the full analysis: Google’s Genie 4: Exploring Its Worlds, Glitches, And Access on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- 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 it as a real-time world model that generates interactive scenes, sound and text-driven changes. The demonstrations are company-selected, public access has not opened, and DeepMind says scenes can drift after a few minutes.
Google DeepMind presented Genie 4 on October 27, showing an AI system that generates interactive scenes as users move through them and enter text instructions. The company says the model can create visuals and audio together, but access is currently restricted to trusted testers, and DeepMind acknowledges that details can drift during sessions lasting only a few minutes.
In the demonstrations, users steer a generated scene with keyboard or mouse controls and can type changes while it runs, such as changing the weather or adding an object. DeepMind describes Genie 4 as a world model: it predicts what the next moment should look and sound like based on the scene so far and the user’s input, rather than rendering a view from a stored 3D map as a conventional game engine would.
DeepMind says the system generates ambient sound, music and dialogue alongside images, and simulates movement and interactions such as collisions. It also says Gemini 3.8 Flash can interpret goals, break them into steps and monitor the generated scene. These are vendor descriptions illustrated by selected company demonstrations; DeepMind has not published benchmark numbers substantiating the claimed gains in realism or consistency.
The company says Genie 4 is available to trusted testers. It plans to offer access through Project Genie to Google AI Ultra subscribers and to early Gemini API partners, but has announced neither a launch date nor API pricing. DeepMind says generated video and audio carry an invisible SynthID watermark. That does not by itself establish how well the watermark can be detected or retained after content is edited or shared.
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
From Generated Clips to Interactive Worlds
Genie 4 points toward a different use of generative AI: instead of producing a fixed video from a prompt, it attempts to maintain a scene that reacts to controls and instructions. If that works reliably, developers could use such systems to explore game concepts and film previsualization without building every environment by hand. DeepMind reports tester feedback involving those uses, but the examples are not independent evaluations or evidence of broad commercial deployment.
The same capability could matter in robotics research, where simulated settings let teams test tasks without putting a physical machine into every trial. The demonstration described by DeepMind used a simulated Apptronik Apollo humanoid playing checkers; no real robot was involved. DeepMind’s claims about physical behavior therefore should not be read as proof that Genie 4 can safely or accurately train robots in real-world conditions.
For audiences, generated worlds with synchronized sound may make synthetic media more convincing. The stated SynthID marking is a relevant disclosure measure, but the announcement does not provide testing results on watermark robustness or on viewers’ ability to identify generated clips. Those questions remain important as access expands.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with Mic 2, 1 CMOS, 4K/120fps, 3-Axis Stabilization, 128GB Card
As an affiliate, we earn on qualifying purchases.
Genie’s Move Toward Real-Time Scenes
Genie 4 follows a series of research systems from Google DeepMind. The original Genie, introduced in February 2024, generated simple 2D worlds from video training data. Genie 2, presented in December 2024, generated 3D environments from a single image and could maintain them for up to a minute, though real-time operation required a reduced-quality version.
DeepMind presented Genie 3 in August 2025 as its first real-time world model, describing output at 720p and 24 frames per second and consistency for a few minutes. Google later opened Genie 3 through Project Genie to paying U.S. subscribers, according to the source material; TechCrunch reported a 60-second limit per world, citing computing costs. That reported limit is distinct from DeepMind’s public description of the model’s duration.
DeepMind’s description of Genie 4’s “few minutes” of continuous interaction is not a guarantee that every scene remains stable for that long. TechCrunch and Bloomberg, as cited by German publication heise, reported demonstrations lasting about two minutes; that specific duration does not appear in DeepMind’s own account. Genie 4’s improvement over Genie 3 is therefore difficult to compare from public information alone.
“Details and objects can shift or drift over time.”
— Google DeepMind
Access, Benchmarks and Scene Stability
Several practical details are not public. Google has not given a date for wider Project Genie access, stated the Gemini API price or explained how many users will be included in early access. It also has not released benchmark results that would allow outsiders to compare Genie 4’s realism, audio quality, response speed or reliability with other systems.
The degree of scene drift is also unclear. DeepMind says objects may change over time and describes interaction lasting a few minutes, while external reports cited in the source material put demonstrated scenes at about two minutes. Those figures refer to reported demonstrations, not a published universal limit. The available clips were selected by DeepMind, so they do not show how often the system fails across ordinary prompts or longer sessions.
Use cases remain at different stages. Film and game applications are described as tester feedback, while the Apollo example is a research simulation. The announcement does not establish that Genie 4 is ready for production workflows, provides accurate physical simulation for robotics, or offers a way for users to verify SynthID marks after distribution.
Google’s Planned Access Rollout
The next announced step is access for Google AI Ultra subscribers through Project Genie and early partners using the Gemini API. Google has not said when either group will receive access, how broad the rollout will be or whether the API will have usage limits similar to earlier Project Genie access.
Further public demonstrations, technical documentation or independent tests would help clarify whether Genie 4’s scenes remain coherent, how quickly it responds to text instructions, and how its audio performs. Until those details are available, the confirmed development is the announcement and limited testing—not a generally available product or a verified set of performance benchmarks.
Key Questions
What is Google Genie 4?
Genie 4 is an AI world model that generates a scene as it unfolds in response to user controls and text instructions. DeepMind says it creates images and audio together rather than producing only a fixed video clip.
Can the public use Genie 4 now?
No. DeepMind says access is currently limited to trusted testers. The company plans access through Project Genie for Google AI Ultra subscribers and for early Gemini API partners, but has not announced a date.
How long can a Genie 4 world run?
DeepMind describes a few minutes of continuous interaction and warns that objects or details can drift. Reports cited by TechCrunch and Bloomberg, through heise, describe demonstrations of about two minutes; the company has not published a fixed general duration.
Does the robot demonstration show a real robot using Genie 4?
No. The checkers example involving an Apptronik Apollo humanoid was a simulation, according to the source material. It does not show a physical robot performing the task.
Has Google published proof that Genie 4 is more realistic?
DeepMind describes improvements in realism, physics and audio, but the announcement provides no benchmark figures for independent comparison. The public demonstrations were selected by the company.
Source: ThorstenMeyerAI.com
