Carnegie Mellon University
Browse

Generative Image AI for Design

Download (45.69 MB)
thesis
posted on 2023-06-07, 20:51 authored by Chengzhi Zhang

Generative image AI tools have sparked heated discussion in the creative community. How generative visual AI can be used for the design community is understudied. From pilot studies and early explorations, we found that designers find image AI an excellent channel for inspiration search in the early design stage. Using a design sketch as the input could significantly reduce the uncertainty of the outcome. Moreover, the fast representation and high fidelity characteristic of diffusion models’ outcome may help bridge early-stage design communication. Based on these pilot study insights, I conducted two studies followed by the interview session. In study 1, I recruited designer participants and conducted a design sketch-AI collaborative workflow study. In study 2, I recruited designer and client participants and conducted a remote design study with designer-client participant pairs. Study 1 characterized image AI - design sketch collaborative workflow. Study 2 portrayed image AI-supported design communication. Study 1 provides insights about what image AI can or cannot do for design practice; the challenges designers encounter while working with diffusion models, and the improvement directions for image AI to serve as better design tools. Study 2 profiled image AI as a resource for bridging design communication and provided insights for future image AI system construction for bridging design communication. 

Funding

MSCD research fund

History

Date

2023-05-10

Degree Type

  • Master's Thesis

Department

  • Architecture

Degree Name

  • Master of Science in Computational Design (MSCD)

Advisor(s)

Daragh Byrne, Nikolas Martelaro, Paul Pangaro

Usage metrics

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC