Making It Real: Advancing 3D Generative AI Through Fabrication Constraints

Current 3D generative AI tools focus on rapidly generating models with the aesthetic quality users want, based on text or images. This approach facilitates iterative design work and creative exploration.

However, 3D models intended for actual Fabrication must satisfy structural constraints based on physical principles, beyond just aesthetic quality. Current generative AI fails to reflect these requirements, resulting in a limitation where models look excellent externally but are difficult to actually fabricate or fail to function properly.

To bridge the gap between digital creations and the real world, this research proposes a new approach to augmenting 3D generative AI that uses not only aesthetic appearance but also physical properties as constraints. Through this, the study aims to extend the creative potential of generative AI into the tangible realm.