Dreaming in Voxels: How AI is Generating Playable Minecraft Worlds

Researchers from New York University have released Dream-Cubed, a large-scale dataset of Minecraft worlds comprising tens of billions of cubes, and trained a family of transformers to generate fully playable, interactive 3D environments at cube resolution. The models enable targeted inpainting, large-scale outpainting, and user-conditioned generation of infinitely sized worlds with fine-grained block-level control.

Generative AI has conquered images, video, text. But what about interactive 3D environments? We trained models on billions of cubes to generate fully playable, structured worlds We are excited to share our latest work, together with New York University: Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes . Generative AI has made incredible progress in language modeling, far beyond other modalities, where words and tokens offer a natural compositional unit for scalable training. This is similar to Minecraft and many other popular video games, where developers rely on cubes, tiles, and other discrete primitives to build rich, interactive worlds. In this work, we show that using cubes as tokens allows large transformers to do the same. Our contribution is two-fold: - We release Dream-Cubed to the research community, a large-scale dataset of Minecraft worlds designed for generative modeling. Our data comprises tens of billions of carefully-balanced cubes from procedurally generated Minecraft terrain and high-quality human-authored maps obtained with the authors’ consent . - We use our data to train a family of powerful transformers for efficient generation of interactive 3D environments at cube resolution. We show how our models allow players to mold the world around them by generating structures, terrain, and maps that are immediately editable and playable. Using high-quality data, we demonstrate that these models can be successfully trained with different training objectives, including both continuous and discrete diffusion, unlocking targeted inpainting, large-scale outpainting, and user-conditioned generation of infinitely sized worlds with fine-grained block-level control.