Abstract
Architecture has always been a means of communicating stories through its design, with its structures and spaces serving as visual narratives. However, recent advancements in technology have created opportunities for architects to enhance their storytelling capabilities through the use of text-to-image algorithms. These algorithms have the potential to improve visual narratives by enabling architects to translate written descriptions into tangible visual representations. This article explores the architecture of visual narrative and how text-to-image algorithms can enhance it in diverse styles. This inquiry aims to help architectural epistemology understand and foresee the potential impact of this technology on the field of architecture. To understand the limits of AI in generating styles to enhance architectural narrative, six distinct styles were chosen for experimentation. The styles were selected based on their unique features, including an architect’s style, movement, or era. These styles include Zaha Hadid, Brutalist, modern-minimalistic, Peter Zumthor, Gothic, and Gaudi. The narrative was kept the same for each style while observing the changes in AI-generated visuals. The results were evaluated by comparing AI’s interpretations in terms of stylistic, environmental, material, form-based, and atmospheric features. While the results showed promise in terms of variations in each category, AI was not successful in implementing all stylistic features while keeping the narrative stable. In particular, after the second environment layer, the modern-minimalistic, Zumthor, and Brutalist styles lost their distinct features, while Gothic and Gaudi-inspired visuals were hardly generated even in the second environment layer. As a result, AI performed well in generating detailed environmental features without any given narrative and creating an atmospheric environment with enlightening the environment for the last layer.
Get full access to this article
View all access options for this article.