Roblox is stepping into the future of AI-driven 3D modeling with the launch of Cube 3D, its first foundational AI model designed for generating 3D objects. As part of its innovation strategy, Roblox is also open-sourcing the model to enable developers to experiment and enhance their creative workflows.
“With Cube, our goal is to make 3D creation significantly more efficient,” the company stated in a press release. “By leveraging 3D mesh generation, developers can rapidly explore creative directions and enhance productivity by quickly determining which concepts to pursue.”
How Cube 3D Works
Cube 3D is trained specifically on 3D objects, drawing inspiration from modern AI models that predict the next logical step in a sequence. Roblox explains that just as language models learn to anticipate the next word in a sentence, Cube 3D tokenizes 3D objects and predicts the next shape token to assemble a complete model.
According to Samantha Spielman, a Roblox spokesperson speaking with The Verge, the training data includes a mix of licensed datasets, publicly available 3D assets, and experience data gathered from the Roblox ecosystem.
Future Capabilities of Cube 3D
The development of Cube 3D is far from over. Roblox envisions the model evolving into a multimodal AI tool capable of generating objects not only from text but also from images, videos, and other input formats. This aligns with Roblox’s broader AI strategy to integrate Cube 3D with its suite of AI-driven creation tools, further empowering developers and content creators.
With Cube 3D, Roblox is taking a significant step toward democratizing 3D content creation, offering game developers an AI-powered way to bring their visions to life with unprecedented speed and efficiency.
