
AI-generated editorial illustration of an explorable miniature world.
Google is opening Genie 4 to some AI Ultra users. It runs in Project Genie in a web browser. You can explore a world, change it and hear sound.
The November 18 rollout brings the model to more users. DeepMind first showed it to a small test group in October.
How Genie 4 changes a world
Genie 4 builds a world as you move through it. A new prompt can change the weather or add a character. You can also use words to steer the action.
The sound changes with the scene. It can include background noise and speech from characters.
Our first look at Genie 4 covered its October preview. The research overview explains how the model makes these worlds.
An agent helps build a world
The Project Genie update adds a tool to help create scenes. It brings Google's models together to set the look and score. Gemini can watch the scene and suggest where to go next. It can also make an event happen.
A new gallery holds worlds to explore, share and remix. One is a mystery set in a gothic manor. Another lets users see an ant colony from an ant's view.
Genie 3 is still in the model picker. It now offers up to three minutes per run. Users can start from a full-circle image or link worlds together.
Who can try the new tools
Google AI Ultra users who meet the rules can start now. AI Pro users can join a waitlist on the same site.
Google says the models still have limits. The camera can lag. Complex physics, precise sound and lasting scene detail need more work.
The service checks inputs and outputs for safety. It adds SynthID marks to the video and sound it makes.
Our guide to Google covers the firm behind these tools.
Read also
Popular news

OpenAI Signs Six-Year Deal for 457 MW of AMD Compute From TensorWave

When GTA 5 launched, ChatGPT was still nine years away

ChatGPT Can Bypass the “I’m Not a Robot” Test: How Is That Possible?

AI Agent Permissions: What to Allow, Review and Block

Nvidia revenue doubles to $115.8 billion as Rubin ships



