Roblox opened its 4D object generation tool to all creators on February 3, following early access that started in November 2025 with nearly 200 creators. The feature has already produced 1.8 million objects since the underlying Cube 3D model launched in March 2025.
The tool generates interactive objects with moving parts, physics, and behaviors from text prompts. Creators use schemas like Car-5 for drivable vehicles with spinning wheels or Body-1 for single-mesh items. Generation takes 20 to 40 seconds per object and runs through safety filters for both prompts and outputs.
Early access numbers show traction. The Wish Master game generated over 160,000 objects during the test period, with average playtime increasing 64%. The system works through Roblox Studio's GenerateModelAsync API and supports in-experience replication.
This represents an evolution from static 3D object creation. Where Cube 3D handled visual assets, 4D adds interactivity without requiring manual scripting. The company plans to add custom schemas and eventually full scene generation including code and animations.
The timing is notable. The announcement came shortly after Google revealed its Genie tool, which triggered a stock dip for Roblox alongside gaming companies like Nintendo and Take-Two. Whether this is defensive positioning or coincidental timing, the feature extends Roblox's creator tools at a moment when generative AI for game development is heating up.
One practical concern: Roblox is deprecating its original Mesh API in six weeks, requiring creators to migrate to the new system. The company open-sourced the Cube foundation model last year, though the 4D features remain proprietary.
The real test comes when thousands of creators start using this at scale. Early playtime numbers are promising, but production use will reveal whether generated interactive objects hold up across the platform's massive variety of experiences.