Case Study of GAI for Generating Novel Images for Real-World Embroidery

2025-10-17

AI Summary

This study addresses the practical challenges of acquiring culturally resonant patterns and achieving fine-grained customization of details and colors in embroidery art. We propose a generative AI–assisted design method tailored for disabled creators. Through an autoethnographic case study, we established a disability-led design team and integrated constraints from traditional hand embroidery—including stitch types, thread materials, and grid-based precision—to conduct iterative prompt engineering and develop a lightweight, domain-specific GPT model for embroidery pattern generation. Our key contributions are twofold: (1) the first deep adaptation of generative AI to the logical structure of traditional embroidery craftsmanship, and (2) a disability-centered framework for inclusive human–AI co-design. Empirical evaluation confirms the physical embroiderability of AI-generated patterns, significantly enhancing cultural fidelity and accessibility of personalized creation. We further identify critical bottlenecks in output stability and controllability, offering a methodological paradigm and optimization pathways for deploying generative AI in digital heritage preservation.

Technology Category

• Humans and AI: AI for Accessibility
• Intelligent Robots: Embodied AI
• Philosophy and Ethics of AI: Artificial General Intelligence

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

In this paper, we present a case study exploring the potential use of Generative Artificial Intelligence (GAI) to address the real-world need of making the design of embroiderable art patterns more accessible. Through an auto-ethnographic case study by a disabled-led team, we examine the application of GAI as an assistive technology in generating embroidery patterns, addressing the complexity involved in designing culturally-relevant patterns as well as those that meet specific needs regarding detail and color. We detail the iterative process of prompt engineering custom GPTs tailored for producing specific visual outputs, emphasizing the nuances of achieving desirable results that align with real-world embroidery requirements. Our findings underscore the mixed outcomes of employing GAI for producing embroiderable images, from facilitating creativity and inclusion to navigating the unpredictability of AI-generated designs. Future work aims to refine GAI tools we explored for generating embroiderable images to make them more performant and accessible, with the goal of fostering more inclusion in the domains of creativity and making.

Problem

• Using GAI to generate accessible embroidery art patterns
• Applying GAI as assistive technology for embroidery design
• Refining GAI tools for performant embroiderable image generation

Innovation

• Generative AI creates embroidery patterns
• Custom GPTs tailored for visual outputs
• Iterative prompt engineering for embroidery requirements