Cube: A Roblox View of 3D Intelligence
This addresses the problem of automating 3D content creation for Roblox developers, but it is incremental as it focuses on initial tokenization rather than a complete model.
The paper tackles building a foundation model for 3D intelligence to support Roblox developers in generating 3D objects, scenes, and scripts, presenting a 3D shape tokenizer as a first step and demonstrating applications like text-to-shape generation.
Foundation models trained on vast amounts of data have demonstrated remarkable reasoning and generation capabilities in the domains of text, images, audio and video. Our goal at Roblox is to build such a foundation model for 3D intelligence, a model that can support developers in producing all aspects of a Roblox experience, from generating 3D objects and scenes to rigging characters for animation to producing programmatic scripts describing object behaviors. We discuss three key design requirements for such a 3D foundation model and then present our first step towards building such a model. We expect that 3D geometric shapes will be a core data type and describe our solution for 3D shape tokenizer. We show how our tokenization scheme can be used in applications for text-to-shape generation, shape-to-text generation and text-to-scene generation. We demonstrate how these applications can collaborate with existing large language models (LLMs) to perform scene analysis and reasoning. We conclude with a discussion outlining our path to building a fully unified foundation model for 3D intelligence.