When Google DeepMind introduced Genie 3 in the summer of 2025, he placed it in a very specific territory: a world model capable of generate a visual environment y react in real time how you move through that world, designed above all as a tool to train AI agents. Now, that technology takes a (small, but symbolic) step towards the public: Google has launched Project Geniusa prototype that allows external users to “create, explore and remix” interactive worlds generated by the model.
Access, however, comes with clear conditions. To try it you need the Google AI Ultra subscription, which Google announced at $249.99 per month (≈$250) and that, at least at this start, is aimed at profiles that want “the most” from their AI stack. Additionally, Project Genie is limited to USA and requires to be over 18 years old.
Three modes and a “prompt → sketch → world” flow
Project Genie launches with three ways to interact:
- World Sketching: You describe the world and your character, you choose perspective (first person, third person or isometric) and the system shows you a preview before entering. This “sketch layer” is generated by Nano Banana Pro, which allows the prompt to be iterated with quick visual feedback.
- Exploration: You move through the environment and the model “fills” and reacts as you move.
- Remixing: you can start from worlds created by others, modify them and generate new variations.
In other words: Google is trying to package a world model as a quick creative tool, closer to “exploring ideas” than “producing a game”.
It is not an engine, and its limits matter
Here it is worth highlighting a red line that the approach itself reveals: Genie 3 is not a game engine. Its outputs may look “game-like” and simulate physical interactions, but there is no traditional game design layer (systems, progression, rules, persistence, etc.). Furthermore, experience is limited: in Project Genie, generations are limited to 60 secondswith output 720p and around 24 fps.
And yet, it’s easy to understand why this is catching the market’s attention. Financial media have already interpreted it as a sign of ambition in the “tools for creating worlds” space, to the point of generating nervousness around Unity (although today the comparison is more narrative than technical: a world prototype does not replace a production engine).
Why are you interested in gaming and esports?
For gaming, Project Genie is a public demo of the “near future” in preview and prototyping: go from idea to navigable space in minutes. To esportsthe impact is indirect but real: if these models mature, they can make it cheaper to create scenarios, interactive content for viewers, “cross-map” experiences or activations for fans… without the need for complete art and programming pipelines.
In short: Project Genie does not inaugurate a new engine, but it does open a window to a technology that, over time, could change how visual ideas are generated and tested in video games. And the fact that Google is putting it (even behind a $250 wall) in the hands of outside users is, in itself, a sign that it wants to accelerate that conversation.