Interacting with Architectural Generative Design Models Using Language Models

This project explores the potential of large language models (LLMs) to facilitate novel workflows with generative design algorithms, enabling the intuitive creation of architectural designs.

A typical parametric or computational design approach demands that users input precise numerical or categorical data, which can be complex and tedious. Additionally, computational analysis often yields false color images or numerical spreadsheets, which can be challenging to interpret in light of the project's objectives.

Can the ability of large language models to bridge natural language and structured data simplify and enhance the interaction between architects and design algorithms?

The research will assess the effectiveness of large language models in generating and interpreting structured data for design models. Through prototypical case studies, the project aims to discover new possibilities in building massing, floor plan layout, and facade patterns. The expected outcomes include a series of short videos and an overview paper, providing insights for integrating natural language with computational design tools.
 

figure

Funded by: ETH Career Seed Award
Duration: Oct 2023- Sept 2024

Principle Investigators:

Researchers:

JavaScript has been disabled in your browser