Google DeepMind Opens Project Genie, Its AI World-Generation Tool, to Ultra Subscribers

Google DeepMind is opening up early access to Project Genie, an experimental tool that generates short, interactive game-like worlds from text prompts or images, starting Thursday for Google AI Ultra subscribers in the U.S. The prototype combines DeepMind’s latest world model Genie 3 with its image generator Nano Banana Pro and Gemini, letting users create explorable environments in seconds.

The release comes about five months after Genie 3’s research preview and is aimed at expanding testing as DeepMind develops more capable “world models” — AI systems designed to simulate environments and predict how they change over time. While researchers see world models as an important long-term area of AI progress, DeepMind’s near-term focus is entertainment, using video game-style experiences as an early application.

Project Genie works through a “world sketch” process. Users describe a setting and main character with prompts, Nano Banana Pro produces an initial image, and Genie builds an interactive world from that starting point. People can also remix existing worlds, browse curated examples, and download videos of what they explore. DeepMind says results can be inconsistent: the system often excels with whimsical or artistic styles but struggles with photorealistic scenes, and worlds generated from real photos don’t always match the source closely.

DeepMind is also keeping the experience tightly limited. Sessions currently allow only 60 seconds of world generation and navigation, reflecting the heavy compute requirements of Genie 3’s auto-regressive design. The company says the cap helps make the prototype available to more users while it remains in an early stage.

Researchers are upfront that Project Genie isn’t a polished product yet. Navigation can feel clunky, characters may clip through objects, and interactions are still relatively simple. But by widening access through its premium subscription tier, DeepMind is inviting more feedback as competition around world models grows, with other labs and startups also pushing into interactive simulation.

For now, Project Genie offers a glimpse of how AI-generated environments could evolve — impressive in its creativity, limited in realism, and very much still a work in progress.

See more here:

This analysis is based on reporting from TechCrunch.

Image courtesy of Google.

This article was generated with AI assistance and reviewed for accuracy and quality.