AI & Animation Education

Home  /  Knowledge Base  /  CLAY: A Controllable Large-scale Generative Model for Creating High-quality 3D Assets

Research · Technical

CLAY: A Controllable Large-scale Generative Model for Creating High-quality 3D Assets

ACM SIGGRAPH 2024 · Jul 2024

Key points

  1. A 1.5 billion parameter native 3D generative model with multimodal conditioning.
  2. Received a SIGGRAPH 2024 best paper honourable mention.
  3. Open sourced, making large scale 3D generation directly usable in teaching labs.

Summary

CLAY is a 1.5 billion parameter generative model designed for native 3D asset creation with multimodal conditioning, presented at SIGGRAPH 2024 where it received a best paper honourable mention. Unlike 2D-to-3D lifting approaches, CLAY operates directly in 3D space and accepts conditioning from multiple input modalities. The model is open-sourced, making large-scale 3D generation practically accessible for teaching labs. It represents the capability milestone against which game-art and 3D animation programs are now calibrating their curriculum treatment of AI asset generation.

Source

Source: ACM SIGGRAPH 2024 ↗ (Research)

Cite this item

ACM SIGGRAPH 2024 (2024). ‘CLAY: A Controllable Large-scale Generative Model for Creating High-quality 3D Assets’, ACM SIGGRAPH 2024. Available at: https://dl.acm.org/doi/10.1145/3658146

Download .risDownload .bib

Your reference manager can also read this page directly: with the Zotero (or Mendeley) browser connector installed, save it straight to your library. Whole-collection exports: RIS, BibTeX, CSL-JSON.