CreativeFlow boosts creative variety in 3D asset generation
CreativeFlow: A One-to-Many Analogical Relation Transfer Method for 3D Asset Generation
Artificial IntelligenceGraphics
Summary
Creating 3D objects from text often results in similar-looking models, limiting creativity. The authors introduce CreativeFlow, a method inspired by how people think analogically to produce various related but distinct 3D designs from the same text. Their approach pairs source and target assets that look connected but are geometrically unique. Experts found that CreativeFlow produces more novel and visually interesting 3D models. This work also provides a new dataset and benchmark to help train future 3D model creation systems.
What this means in practice
- •For 3d artists and designers: Create a wider variety of distinctive 3D assets from textual descriptions for games, movies, or VR experiences.
- •For 3d software developers: Incorporate relation-aware analogical generation to improve creativity and diversity in text-driven 3D model tools.
Authors
Xuechen Li, Shuai Zhang, Nanxuan Zhao, Qing Chen
Abstract
Inspired by cognitive science, we present CREATIVEFLOW, an analogical generation framework that explicitly models analogical divergent thinking to mitigate creative homogenization in text-to-3D pipelines. Our method derives a series of meaningful yet relationally similar source-target asset pairs, each featuring distinct geometric configurations. Expert evaluations demonstrate that our framework substantially enhances creative novelty and visual fascination. This workflow and its resulting assets establish a foundational dataset and benchmark for future relation-aware 3D model training.