# Google’s Genie 4 Lets You Step Into an AI World—and Change It as You Go

> Project Genie’s November 18 update adds real-time world editing, native audio and a community gallery for eligible Google AI Ultra subscribers.

By Lidia Yadlos · November 19, 2026

Canonical: https://blockster.com/google-genie-4-interactive-worlds-ultra-launch

Google's latest creative AI release is something you can move around inside. [Genie 4 is rolling out through Project Genie](https://blog.google/innovation-and-ai/models-and-research/google-deepmind/genie-4-now-available-project-genie/) to eligible Google AI Ultra subscribers, following its earlier research preview.

The November 18 update lets users alter a generated environment while exploring it: change the weather, introduce a character or redirect the action with a new prompt. Sound is generated alongside the scene, including ambient audio and dialogue.

## A world that responds instead of a video that plays

The distinction is interaction. A generated video follows a sequence; a world model responds to what the user does next. That makes the format interesting for creative prototypes, experiments and experiences that sit somewhere between a film and a game.

The update also introduces a Project Genie agent to help construct worlds conversationally, plus a gallery for discovering, sharing and remixing them. Google AI Pro members can join a waitlist rather than assuming immediate access.

These features should not be confused with a complete game-development engine. A compelling moment inside a generated scene is only one component of a reliable, repeatable interactive product.

## The older model gets a different upgrade

Google is keeping Genie 3 in the model picker. Its update includes sessions of up to three minutes per generation, panorama-based starting environments and linked worlds. The three-minute figure belongs to Genie 3, not a newly announced Genie 4 duration.

> “Genie 4 remains an early, evolving area of research.”
>
> — Google DeepMind, Genie 4 research overview

Google's [research overview](https://deepmind.google/blog/genie-4-a-new-generation-of-world-simulation/) acknowledges problems with long-term consistency, object permanence and complex physical interactions. A door, object or spatial relationship can drift as a session continues.

That qualification makes the practical use clearer: explore an idea, test a mood or develop an interactive concept, while checking what the system actually preserves. Alongside the builder tools in Blockster's [Gemini Hydrogen coverage](https://blockster.com/google-gemini-4-hydrogen-search-ai-builders), Genie shows Google pursuing both sides of creation—the software used to build experiences and the experiences generated directly.
