Depicting Wood: Mass Timber Diffusion Models

In the outskirts of the Texas Triangle Megaregion, East Texas boasts an extensive 12 million acres of sustainable and productive forest land. A comprehensive forest inventory and analysis reveal an annual Southern Yellow Pine (SYP) timber volume in this region that surpasses industry harvest demands by an impressive 90 percent. Remarkably, this rapidly growing and regenerative timber species has the potential to yield a thirty-ton truck load every nineteen seconds. Despite the profound environmental, economic, and social impact on both rural and urban Texas communities, particularly within the design, manufacturing, and construction sectors, the majority of upcoming building projects overlook the prospect of incorporating locally sourced mass timber systems. The architectural challenges at hand are twofold—technical and conceptual—highlighting gaps in awareness and understanding. This oversight prevents the exploration of an opportunity that could significantly contribute to circularity, resiliency, and an equitable construction culture in the region.To address these barriers and harness the potential of utilizing locally harvested mass and cross-laminated timber (CLT), we employ Artificial Intelligence (AI), specifically Diffusion Models (DMs), a text-to-image workflow. This innovative approach aims to visualize and generate targeted architectural typologies and materiality efficiently and productively. Beyond their advantages in terms of speed and fidelity, this research explores how text prompts can be optimized to depict architectural designs through the lens of AI.The study establishes a comprehensive survey of keywords and text prompts through a method of progressive prompt engineering. The objective is to create a benchmark and reference for teaching, research, and practice by identifying successful keywords specific to timber construction and housing typologies. Simultaneously, this research introduces an efficient process to enhance awareness and applicability of timber in the design phase. Furthermore, the method of progressive prompt engineering provides a high degree of control in developing unique formal architectural elements and strategies, essentially establishing a distinct “wood language” within DMs.

https://doi.org/10.35483/ACSA.AIA.InterMaterialEco.23.24

Volume Editors
Caryn Brause & Chris Flint Chatto