Google opens up Genie 4: you can change the world with text on the fly, but a session lasts only a minute
Starting November 18, Google is opening up Genie 4, a neural network that builds an interactive world while you walk through it. During a session you can type "let it rain" or "add a passerby," and the scene rebuilds itself along with its audio, including characters' lines. But Genie 4 currently runs continuously for only about a minute, and access is open to Ultra subscribers. For game and simulation teams, this is a reason to sketch a location prototype, not to build a product.
AI-processed from TG: @neyr0graph (NeuroGraph); edited by Hamidun News
Ordinary image and video generators follow a "write a prompt, wait, get the finished result" pattern: you cannot intervene in the process. Google's Genie 4 breaks that pattern. It is a model that generates an interactive world while you walk through it, and it lets you change that world with text commands on the fly.
Imagine walking down a street drawn by a neural network, typing "let it start raining," and watching the sky darken as the street noise gives way to the sound of a downpour. The main caveat from the headline: a continuous Genie 4 session currently lasts only about a minute, so it is closer to a quick sketch than a finished game environment. Worth reading for anyone building games, simulations, or generative products and deciding whether to try this format now.
What's new in Genie 4
On November 18, Google began rolling out Genie 4 in Project Genie. There are two main updates. The first is control over what happens through text commands during a session: you can ask it to change the weather, add a character, or steer the action in a different direction without restarting the scene. The second is real-time audio generation. The environment rebuilds itself, and the sound changes with it, down to characters' dialogue.
Alongside the model, Google released several companion features:
- Project Genie agent — an assistant that creates worlds through conversation. It works with custom soundtracks and Gemini prompts.
- Community gallery — a place to explore other people's worlds and create your own versions.
- Genie 3 updates — sessions of up to three minutes, connecting worlds, and building environments from 360° panoramas.
Honest limitations
The most important detail is easy to miss behind the announcement: Genie 4 currently offers about a minute of continuous interaction. Three-minute sessions apply to Genie 3. In other words, the newer model, with live audio and on-the-fly edits, currently runs shorter than its predecessor.
Google itself points to weak spots: accuracy in reproducing real places, text rendering, and audio. If you expect an exact copy of a specific street or readable signs in a generated world, you should not count on that yet.
Where to use it and how much it costs
Project Genie runs in the browser. It requires a personal Google account and age 18 or older. Google says the service is available in more than 140 countries, but daily limits apply, and they depend on the plan and on load.
Here is how the plans break down. For Google AI Ultra subscribers, access to Project Genie is provided without a waitlist: once you sign in to your account, the service should open. Google AI Pro, which costs $19.99 per month in the US, has a waitlist. That subscription lets you join the queue; it does not give immediate access to Genie 4.
What to do with this in practice
For developers of games, simulations, and training environments, Genie 4 is for now useful as a rapid prototyping tool. A minute is enough to test an idea: how a location looks, how a scene sounds, how the world reacts to commands. It is not enough for a full game or a long simulation.
Product teams should look at the pattern itself: steering generation through conversation as it happens, rather than with a single prompt at the start. It can carry over to other services where the result is built gradually and users want to step in along the way.
The community gallery shows another familiar mechanic: take someone else's world and make your own version. Such remixes usually help grow an audience around a generative tool quickly.
Bottom line: Genie 4 shows where generative models are heading — from "order and receive" to "be inside and adjust." For now it is an experiment limited in time and access, but you can try it right in the browser if you have an Ultra subscription.
- Nov 20, 2026Anthropic raises $81.3 billion in IPO, beating SpaceX's $75 billion record
- Nov 20, 2026PromptFight arena: an agent without the code from the rules gets a 403, and after 10 errors, a ban
- Nov 20, 2026Google makes its Gemini Spark agent free and renames it Tasks
- Nov 20, 2026Mistral releases Vibe desktop app with parallel agent tasks
Key Questions Answered
What's new in Genie 4?
On November 18, Google began rolling out Genie 4 in Project Genie. There are two main updates. The first is control over what happens through text commands during a session: you can ask it to change the weather, add a character, or steer the action in a different direction without restarting the scene. The second is real-time audio generation. The environment rebuilds itself, and the sound changes with it, down to characters' dialogue.
What are the key details on Honest limitations?
The most important detail is easy to miss behind the announcement: Genie 4 currently offers about a minute of continuous interaction. Three-minute sessions apply to Genie 3. In other words, the newer model, with live audio and on-the-fly edits, currently runs shorter than its predecessor. Google itself points to weak spots: accuracy in reproducing real places, text rendering, and audio. If you expect an exact copy of a specific street or readable signs in a generated world, you should not count on that yet.
Where to use it and how much it costs?
Project Genie runs in the browser. It requires a personal Google account and age 18 or older. Google says the service is available in more than 140 countries, but daily limits apply, and they depend on the plan and on load. Here is how the plans break down. For Google AI Ultra subscribers, access to Project Genie is provided without a waitlist: once you sign in to your account, the service should open. Google AI Pro, which costs $19.99 per month in the US, has a waitlist.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.