Another week and another new generative AI tool is being described as “game over” for developers. Google AI Ultra subscribers in the US now have access to Project Genie, an interactive world creation tool that lets you generate “playable” environments from text prompts, and there appear to be few controls over what it can do.
Early users have already been generating worlds in the style of Nintendo franchises like Mario and Zelda. Some think they’re going to be able to generate GTA VI before Rockstar gets round to releasing it, but is Project Genie really usable as game development software?
What is Google Project Genie?
Project Genie | Experimenting with infinite interactive worlds – YouTube
Watch On
Project Genie is an experimental prototype AI model that allows the use of both text and visual prompts to generate an explorable world and a character to explore it. First up, Google’s image generator Nano Banana Pro conjures up an image preview. This can be adjusted as desired. The Genie 3 world model then generates an environment in real-time as the ‘player’ moves through it.
You may like
You can move in different ways – walking, running, swimming, flying or even soaring over Hyrule hanging from a glider, as demonstrated by the Verge’s Jay Peters in the post below. There’s no map pre-loaded in memory. Genie predicts and renders the path ahead in real time as you explore. And it’s infinite (for 60 seconds).
Google says it’s aim with the roll out is to “learn more about immersive user experiences to advance our research and help us better understand the future of world models”
Google’s new world AI model tool let me generate a bunch of Nintendo-inspired games. Including one featuring Link with a paraglider! Gift link: www.theverge.com/news/869726/…
— @jaypeters.net (@jaypeters.net.bsky.social) 2026-01-31T23:02:54.599Z
At least initially, the model appeared to have few restrictions around the use of copyrighted material. The video above is a fairly convincing approximation of the visual style of The Legend of Zelda: Breath of the Wild. The Verge was also able to generate worlds in the style of Super Mario 64, and other users have been posting clips showing worlds in the styles of Harry Potter and Grand Theft Auto.
It seems that some IPs have since been blocked. Trying to generate Mario-inspired worlds reportedly now results in a warning message about the “interests of third-party content providers.” The Verge was told by Google Deepmind’s product manager Diego Rivas that “as with all experiments, we are monitoring closely and listening to user feedback.”
I’m speechless.I just made my first AI video game using Google Genie 3!Prompt: “make a game that is better than GTA VI, made by Leslie Benzies”GTA VI is so cooked, this is the future of gaming! pic.twitter.com/sLk9LRoNLzJanuary 31, 2026
Will Project Genie the end of video game studios? And will someone generate GTA VI before Rockstar can release it?
Only if people are happy to play a game where there’s not much to do but look around. Google’s claim that Genie can generate “playable worlds” is a bit of an exaggeration. There’s a chasmic difference between a game environment and a game – and between a “movable” character and a “playable” one.
someone made a greenland version of gta 6 using google genie 3 pic.twitter.com/iMqDF4b6uuJanuary 30, 2026
1st person Harry Potter world Via Project Genie by @GoogleDeepMind $Genie pic.twitter.com/lz0FuOuvskJanuary 30, 2026
As one person comments on Google Deepmind’s post on X. “Imagine all the vapid, empty open worlds with gameplay equivalent to superman 64 this could make”. Although Superman 64 did at least have gameplay.
You may like
Project Genie is currently can currently only generate for 60 seconds, but then a world with nothing to do in it isn’t likely to sustain interest for any longer. It’s also limited to a resolution and frame rate of 720p and 24fps, so you’re not going anywhere fast.
Google’s approach to Genie seems to be the same to what we’ve seen before from Meta, X and Google itself: rush it out as an experiment and then start removing things when people complain. It might soon be getting letters from Nintendo and others, but it can hardly be surprised.
When OpenAI launched the Sora 2 AI video generator in October, copyrighted characters were allowed until threats of legal action arrived. Sam Altman expressed an implausible naivety, claiming to have been surprised that brands didn’t want people to be able to create racist AI slop with their IPs.
There’s a possibility the approach is intended to get attention and potentially broker deals. Two months after the Sora 2 controversy, OpenAI and Disney got in bed together with a plan to allow Disney+ viewers to generate videos with the company’s characters.
Project Genie is a prototype web app powered by Genie 3, Nano Banana Pro + Gemini that lets you create your own interactive worlds. I’ve been playing around with it a bit and it’s…out of this world:)Rolling out now for US Ultra subscribers. pic.twitter.com/rNDXn3VUF6January 29, 2026
Pushing the limits of Project Genie from @GoogleDeepMind Trying to break it with crazy worlds and characters. Super impressive how each world has it’s own logic, and each character its own movement and behavior. pic.twitter.com/8KkyxvjYLYJanuary 31, 2026
Plenty of people believe that auto-generated games will be the future someday, but I’m still not convinced it’s a future players will want. Even if people can generate their own games, are they going to be any good? And will everyone want to play their own game that nobody else is playing?
Project Genie is only available to over 18s in the US who subscribe to Google’s most expensive AI subscription plan, Google AI Ultra. That costs $124.99 per month when signing up for a three-month package.