A New Soul of Architecture: AI-directed drawing

This study explores the integration of generative artificial intelligence into architectural creative processes, examining how its inherent stochasticity can be oriented and structured within defined design trajectories. Through the adoption of nodal algorithms, the research develops a methodology to balance computational autonomy with design control, establishing the algorithmic curator as a critical mediator between human intentionality and machine-generated potential. Therefore, nodal algorithms are examined not merely as instruments of automation or stylistic emulation, but as operational frameworks that foreground co-authorship between human and machine, emphasizing control strategies rooted in semantic and procedural constraints. Within the ComfyUI framework, the nodal algorithm architecture organizes the creative process into phases that are discrete yet dynamically linked, with each node serving multiple functions as a control point for design parameters, a filter for semantic and stylistic coherence, and a bridge between human input and machine interpretation. The system incorporates DreamBooth-trained models fine-tuned on architectural datasets, allowing for specialized generation capabilities while maintaining the flexibility needed for creative exploration. Applied to the complex case study of historic urban centers, this approach demonstrates how AI-assisted design can operate within culturally sensitive contexts. The methodology transforms architectural representation into an iterative dialogue between designer and algorithm, where human expertise guides the system's stochastic processes toward meaningful design solutions. Through this co-creative process, the research redefines digital-age authorship while preserving the expressive richness of AI generation, ensuring alignment with design intent through dynamic constraints, and producing hybrid design languages that meaningfully bridge historical context with contemporary innovation.