Generative AI-powered parametric modeling and BIM for architectural design and visualization

The paper presents a new framework that integrates conversational AI, particularly ChatGPT, into established parametric modeling and BIM tools to streamline architectural design. It addresses the challenges of traditional parametric design, which often relies on extensive scripting, by enabling designers to use natural language prompts to generate Python scripts or Grasshopper definitions. This allows for iterative refinement of geometry, materials, and forms without requiring deep coding knowledge. The framework explores two primary workflows: a user-driven approach integrating ChatGPT with Revit and Dynamo, and an AI-driven approach using ChatGPT API with Grasshopper. AI-based visualization tools, such as Veras, are also incorporated to provide immediate visual feedback, significantly accelerating the design feedback loop from concept to refinement. This approach aims to expand architects' autonomy by allowing them to control building forms and components through natural language, fostering a more intuitive and powerful method for conveying design intent. While highlighting advancements in design efficiency and exploration, the study also acknowledges limitations of current generative AI tools, particularly in handling complex or highly customized designs crucial for architectural practice. It also points out the need to consider the physical feasibility of AI-generated designs, as some visually impressive outputs might be structurally impossible. Addressing these challenges is considered essential for fully realizing the potential of AI in architectural design.