Generative AI-Based Form Exploration of Bugis Traditional Architecture: From Vernacular Morphology to Digital Reconstruction

Abstract

This study investigates how generative artificial intelligence (AI) interprets, reconstructs, and transforms the vernacular morphology of Bugis traditional architecture. Using a qualitative computational mixed-method design, the research combines field documentation of twenty Bugis houses with image-based generative modeling using diffusion frameworks. The results reveal that AI is capable of reproducing prominent morphological attributes including the saddle-shaped roof, elevated stilt structure, symmetrical façade, and longitudinal spatial axis demonstrating its ability to learn visually salient patterns from architectural datasets. However, significant transformations emerge in elevation, materiality, and spatial orientation, driven largely by algorithmic biases toward modernist forms. More critically, symbolic and cosmological elements central to Bugis architectural ontology such as vertical spatial hierarchy, ritual thresholds, and carved motifs are frequently diminished or omitted, indicating AI’s limited capacity to interpret intangible cultural meaning. Despite these limitations, AI-generated hybrid forms provide valuable speculative insights into potential vernacular futures, suggesting that generative models can function as creative tools when guided by cultural expertise. The study concludes that generative AI offers meaningful opportunities for visualization and design innovation but must be integrated with culturally informed frameworks to ensure ethical and accurate representation of indigenous architectural heritage.