AIDeveloping

Google unveils Genie 4 world model with real-time edits and audio

New version generates two-minute interactive experiences, uses Gemini 3.8 Flash for simulation

By LineZotpaper AI staff

Published27 October 2026

Updated27 October 2026

Read Time1 min

Sources5 outlets

Reactions2 on X

Follow this storyHear when it moves

Google announced Genie 4, the latest iteration of its AI world model, on Tuesday, touting improvements in realism, real-time editing via natural language, and the addition of audio generation. The model, built on Gemini 3.8 Flash, can create interactive experiences up to two minutes long, double the duration of its predecessor.

Google DeepMind has released Genie 4, the newest version of its world model for generating interactive AI-powered experiences. According to the company, the model introduces four main updates: more realistic worlds, the ability to modify an experience in real time using natural language prompts, audio generation, and the integration of Gemini 3.8 Flash to drive simulations.

Genie 4 can produce experiences that last about two minutes, up from one minute in the previous Genie 3 model. Google is positioning the tool for pre-visualisation in filmmaking, prototyping game ideas, and creating training scenarios for robots. The announcement did not include details on public availability or specific release timelines.

RSS

The Daily Edition

The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.

How it's landing

Ethan Mollick @emollick

Wharton professor

Genie 4 is quite an insane experience, generating game-like enviornments as you go. It isn't flawless but I can type "Neogothic brutalist city inhabited by robots and horses" & it creates one in a minute. Then I asked it to add dragons. Then a friendly otter NPC. All in realtime

27 Oct, 16:37 UTC View on X

Posts on X from people the desk follows.

Analysis

Why This Matters

Advances in AI world models bring closer the possibility of generating dynamic, interactive environments for creative and industrial use without manual rendering.
Real-time language-based editing lowers the barrier for non-technical users to iterate on simulations, with potential applications in education, design, and entertainment.
The integration of Gemini 3.8 Flash suggests a growing convergence between large language models and world simulation, a trend that could shape future AI development.

Background

Genie is a series of AI world models from Google DeepMind that generate interactive 3D-like experiences from prompts. Genie 3, released earlier, offered one-minute simulations. The new model extends that to two minutes and adds audio capabilities, while also enabling real-time modifications via natural language. The underlying use of Gemini 3.8 Flash represents a shift toward combining language understanding with environment generation.

What to Watch

Whether Google releases a public demo or API access, which would indicate readiness for broader adoption.
Adoption signals from game studios, film pre-visualisation houses, and robotics labs, key use cases cited by Google.
Responses from competitors such as OpenAI and Meta, which are also developing world models and simulation tools.

Sources

Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering · TechCrunch
Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering · TechCrunch
Google’s next Genie AI model generates more realistic experiences · The Verge
Genie 4: A new generation of world simulation · Google DeepMind
DeepMind unveils Genie 4 world model promising more realistic physics and humans · Ars Technica

Checked

9 claims in this story checked against 1 source: 4 supported.

Show each claim

SupportedGoogle unveiled Genie 4, a world model with real-time editing and audio generation.The Verge: “Google announced today its Genie 4 world model for generating interactive experiences using AI. The company says the model has four main updates: It allows for more realistic worlds, lets users update an experience in real time with a natural language prompt, can generate audio”
Partly supportedGenie 4 generates experiences that last two minutes.The Verge: “It also can create experiences that are about two minutes long, up from one minute for the previous Genie 3 model.”
SupportedGenie 4 uses Gemini 3.8 Flash for simulation.The Verge: “uses Gemini 3.8 Flash under the hood to help with its simulations.”
Not in the sourcesGoogle announced Genie 4 on Tuesday.The Verge: “Google announced today its Genie 4 world model for generating interactive experiences using AI.”
Partly supportedGoogle DeepMind has released Genie 4.The Verge: “Google announced today its Genie 4 world model for generating interactive experiences using AI.”
SupportedThe company identifies four main updates: more realistic worlds, real-time natural-language updates, audio generation, and Gemini 3.8 Flash for simulations.The Verge: “The company says the model has four main updates: It allows for more realistic worlds, lets users update an experience in real time with a natural language prompt, can generate audio, and uses Gemini 3.8 Flash under the hood to help with its simulations.”
Partly supportedGenie 4 can produce experiences up to two minutes long, double the duration of Genie 3.The Verge: “It also can create experiences that are about two minutes long, up from one minute for the previous Genie 3 model.”
SupportedGoogle is positioning Genie 4 for pre-visualisation in filmmaking, game prototyping, and robot training scenarios.The Verge: “Google is pitching Genie 4 as a tool that could be helpful for things like pre-visualization for filmmakers, prototyping game ideas, and making training scenarios for robots.”
Not in the sourcesThe announcement gave no details on public availability or specific release timelines.The Verge: “But the … Read the full story at The Verge.”

Discussion elsewhere

Hacker News298 points · 164 comments →

Looks cool, but it's very hard to get me excited about yet another proprietary model for which absolutely zero technical details are released (not even the number of parameters), which is safetymaxxed to the limit, and whose usage quota will (most likely) be so limited that it'l…
kouteiheika on Hacker News

I would like to see how it breaks, don't spend all your focus on how nice it looks with specially curated shots. At least give a few examples of what happens when it breaks from running too long. They mention that it can fail under "Understanding current limitations" but don't s…
TheAceOfHearts on Hacker News