Roblox introduces a tool that enables players to use AI to create interactive in-game objects, with a promise of "full scene generation" in the future

Roblox has introduced a new beta feature that enhances its 3D asset generation tool, which was disclosed last year. This tool allows users to create interactive 3D models from text descriptions. The announcement was made via a blog post.

The release of this tool comes on the heels of investor enthusiasm over Google's introduction of the Genie tool for generating worlds, which led to a decline in Roblox's stock value, along with other major gaming firms like Nintendo and Take-Two.

This feature is now available for creators on the Roblox platform, enabling players to create interactive objects from text prompts. Currently, the tool supports two "schemas": a four-wheeled drivable car and a single-mesh object. A demonstration video showed items like cars, a plane, and a gumball-firing gun being created using text inputs.

Players can create items through text prompts when this feature is used in games. | Image credit: Roblox

In practical application, the tool has already seen success in the game Wish Master, where it was utilized to create over 160,000 objects in half a year. According to the game's creator, Laksh, this functionality led to a 64% increase in average playtime.

Anupam Singh, Roblox's Senior Vice President of Engineering, mentioned the ongoing work towards a comprehensive vocabulary schema system that supports any schema and anticipates a future where creators can build any kind of 4D object or action, though no specific timeline was disclosed.

Singh also highlighted an internal initiative known as "real-time dreaming," aimed at fostering new types of experiences. CEO Dave Baszucki demonstrated this concept in a recent social media video, which showed a basic, low-framerate environment akin to Skyrim, shifting scenes from a village to a boat at sea through text-induced changes like a tsunami.

"Our platform's next frontier is driven by AI advancements, allowing creators to craft immersive environments, refine, troubleshoot, and collaborate through natural language prompts," Singh stated. "If it can be imagined, it should be possible to create it."

Following the unveiling of Google’s Genie tool, publishers have responded by emphasizing their own adoption of AI, although there has been skepticism regarding Genie's actual capabilities. Addressing analyst concerns after releasing Take-Two's quarterly results, Take-Two President Karl Slatoff remarked in an earnings call that although Genie is "very exciting," it resembles "procedurally generated interactive videos" rather than a full game engine, noting that game development involves much more than just creating worlds.