Last Thursday, Google introduced an exciting new experimental AI tool known as Project Genie. By Friday, video game stocks dropped significantly. Gaming giants such as Unity Software, Roblox, Take-Two, and AppLovin faced the brunt of this decline on Wall Street.
Currently accessible only to subscribers of Google’s $249 per month AI Ultra plan, Project Genie is a groundbreaking generative AI world model developed by the company’s DeepMind research team. This tool empowers users to craft interactive virtual environments using merely text and image prompts, allowing for the creation of not just settings but also characters that interact authentically within the virtual realm.
Upon exploring several examples of Project Genie in action, it becomes clear why investors enthused about AI would find this tool’s potential impactful for game developers.
Users on Reddit and X are enthusiastically sharing examples of their favorite virtual worlds created with Project Genie:
This Tweet is currently unavailable. It might be loading or has been removed.
This Tweet is currently unavailable. It might be loading or has been removed.
This Tweet is currently unavailable. It might be loading or has been removed.
This Tweet is currently unavailable. It might be loading or has been removed.
This Tweet is currently unavailable. It might be loading or has been removed.
Riley Goodside, a prompt engineer at Google DeepMind, shared a video displaying a 3D box of cigarettes that comes to life as he manipulates it in a subway station setting. Goodside explained that he supplied the Genie 3 model with the prompts “34th Street-Penn Station” to establish the environment and “discarded pack of cigarettes” for the character, along with an initial frame created using Nano Banana Pro.
Mashable Light Speed
This Tweet is currently unavailable. It might be loading or has been removed.
Goodside noted that while users have considerable control over the elements generated through the initial image, they have limited influence over the virtual environment as their character navigates through it.
Some limitations are evident in other examples, such as this representation of a character striving to look in a mirror.
This Tweet is currently unavailable. It might be loading or has been removed.
Moreover, Project Genie appears to treat secondary characters as inanimate objects, as shown in some additional examples shared on X.
This Tweet is currently unavailable. It might be loading or has been removed.
This Tweet is currently unavailable. It might be loading or has been removed.
Consistent with other generative AI tools, there are also indications of potential copyright challenges with some of the generated content.
This Tweet is currently unavailable. It might be loading or has been removed.
However, industry leaders such as Unity’s CEO Matthew Bromberg express confidence that world models will not replace traditional game engines. Bromberg argued that these new tools will complement the expertise of seasoned game developers.
This Tweet is currently unavailable. It might be loading or has been removed.
Even though Project Genie is still in its early stages, its capability to generate intricate “worlds” is already noteworthy.
Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
Topics
Artificial Intelligence
