From concept to space: a new perspective on AIGC-involved attribute translation

Abstract

Drawn inspiration from phenomenal attributes and translating them into heuristic model tools is one of the effective means to promote architectural form innovation. However, over-reliance on perception indicates greater risks in decision-making. Nowadays, AI-generated Content (AIGC) technology combines the advantages of information comprehensiveness and modelling efficiency, providing new possibilities for the translation of architectural attributes. Based on attribute study, this paper proposes a new approach to spatial translation that uses the Generative Adversarial Network (VQGAN + CLIP) to realize the visualization of abstract concepts and then adds multi-dimensional influence through the Keyframe Style Transfer technology. The eclectic attribute is used as an example for the 2D and virtual 3D translation feasibility experiments. The article aims to improve the scientificity and influence of spatial translation through a technical organization from the perspective of architects. While providing an innovative, democratic and efficient aided-design tool also highlights a new angle for AIGC-involved pre-design.