Yoon, J., Jo, GI. (2026). Bridging Imagination and Tectonics: A GAI Workflow for Timber-Structured Library Integrating Children’s Participatory Design. In: Tedjosaputro, M., Lombardi, D., Di Marco, G., Schnabel, M.A. (eds) From Inspiration to Solutions. xArch 2026. Springer, Singapore. https://doi.org/10.1007/978-981-92-3527-8_17
This study examines a workflow for integrating children’s participatory design with AI-driven timber structure visualisation, repositioning Generative Artificial Intelligence (GAI) from an ideation tool to a design mediator. To translate non-expert ideas into technically feasible architectural proposals, a multi-stage GAI workflow was developed and evaluated. The pipeline refines participatory sketches to foreground architectural elements, preceding high-resolution rendering to enrich spatial depth and material expression. Given the complex tectonic demands of timber construction, a customised diffusion model pipeline was established, combining Stable Diffusion with LoRA, ControlNet, and IP-Adapter. To address the tendency of standard diffusion models to generate structurally implausible outputs, a reasoning-based multimodal Large Language Model (LLM) was incorporated for structural pre-learning—enabling the system to internalise tectonic logic prior to visualisation. Under the bioethical constraints attendant upon involving minors, generated outputs were assessed within a proof-of-concept (PoC) framework through a mixed-methods validation approach, encompassing participant observation and expert heuristic evaluation. Quantitative findings indicate appreciable improvements in both user engagement and structural plausibility, accompanied by a reduction in the frequency of human-in-the-loop interventions required. Building on these outcomes, an intera ctive validation framework is proposed to facilitate expert review and stakeholder engagement, balancing democratic participation with structural integrity.
