AN AI-APPROACH TO THE FUTURE OF CREATING FASHION

This paper explores the potential for generative-AI to contribute to conceptualisation and visualisation of fashion design for purposes of streamlining creative processes and minimising resource-intensive garment prototyping.

The practice-based study introduces a conceptual Gen-AI prototype which exploits text-to-image synthesis and the conversion of hand-drawn sketches into photorealistic images.
Utilising latent diffusion models, with fashion-specific parameters defined by end-users, the work-in-progress prototype produces high fidelity 2D garment renderings and visualisations based on designer insights and user preferences. Designers retain control over the parameters and adjustments which enable them to rapidly explore unlimited design variations. This virtual approach could serve as an alternative to conventional design ideation methods and training-intensive 3D virtual prototyping, to speed up design iteration, reduce material waste and promote sustainable practices in the fashion industry.

Our work to date is at an early stage and shows how the prototype is ‘learning and improving’ its capability to interpret designs accurately. We present early qualitative findings gathered through ongoing interactions and feedback gained from prototype demonstrations with various focus groups.
We discuss limitations and focus on potential future developments which could benefit the fashion industry by creating an intuitive interface, supporting better-informed trend forecasting and realising scalability to improve industry workflows.