Releasing generative AI capabilities for designers

Context and objectives

Recent progress in deep learning makes it possible to generate high-quality images from prompts with online tools such as Midjourney or DALL·E. The opportunities span many sectors, including marketing, gaming and media. The project set out to explore the potential of generative AI to support the creative process in the luxury fashion industry. The Emerton Data team worked on it with data and AI experts as well as graphic designers, the target users.

Outcome

• Experiments and user feedback showed that generative AI can boost creativity while preserving the house's strong artisanal creative DNA.
• They also showed that, to our knowledge, existing tools such as Midjourney do not meet advanced creative needs, which are iterative by nature.

Our approach

Step 1a: map the end-to-end creative process in the fashion industry and understand the approach to creation, which has a strong artisanal component.

Step 1b: list and run concrete experiments with the graphic designers, using advanced generative AI tools and techniques such as Stable Diffusion XL, ControlNet, DreamBooth and fine-tuning, to define ad hoc methodologies and assess feasibility.

Step 2: define an efficient user experience and design mock-ups to ensure smooth adoption and integration into the creative process.

Step 3: build the web application that supports the creative process, onboard the designers to start adoption, and iterate.