Description
This book is an interdisciplinary guide for the application of generative AI in the design industry. It serves as a practical guide for designers to use the latest generative AI technologies in expanding their digital design toolset; at the same time, it explores methods to customize latest genAI models to make it better suited for design applications.
Generative AI technologies have matured to the point where it can be helpful for designers and other creative professionals. However, due to the segregation in disciplinary knowledge, few of them know how to best leverage the power of generative AI in their work; and many of them are constrained by out-of-box solutions online that are not apt at serving designers needs. This book aims to break the disciplinary wall for them and equip them with sufficient knowledge in machine learning and generative modeling so that they can find the correct tools to use and even customize generative models to achieve their own desiredeffects. On the reverse side, machine learning researchers have been producing amazing art generation models which are a nice first step but are still detached from design professionals real-world application scenarios. For the more tech-savvy readers, this book will point out directions as to how state-of-the-art generative AI models can improve in order to bridge the gap between AI research and consumer applications in the design industry.
Introduction.- Part 1: Exploring Existing: Generative AI Tools.- Chapter 1: Generative AI tooling landscape.- Chapter 2: Dall-E and Midjourney: online image generation tools.- Chapter 3: Stable Diffusion: open source generative model.- Chapter 4: Other alternatives Introduction to alternative genAI tools in this space, such as Photoshop GenAI plugin, lookX.ai, NewArc.ai, PromeAi etc.- Chapter 5: Beyond out-of-box solutions: This chapter discusses the limitation of out-of-box generative AI solutions, and motivates Part 2 where customization techniques are introduced to solve these issues.- Part 2: Customizing Generative AI Models for Design.- Chapter 6: Model fine-tuning basics using pytorch.- Chapter 7: ControlNet: controlled image editing.- Chapter 8: LoRA for redefining semantics.- Chapter 9: InstructPix2Pix for precisely targeted image editing.- Chapter 10: Other technologies.
Lezhi Li is a machine learning researcher working on large language models and large image generation models. Her career path spans multiple top-tier companies in Silicon Valley. The unconventional start of her AI career began at Harvard Graduate School of Design, where she trained models to evaluate the aesthetics of cityscapes -- among the earliest efforts at the school to integrate AI into design research. From multidisciplinary design training to advanced work in foundation models, her career reflects a sustained convergence of creativity and scientific analytical thinking. As AI increasingly serves as a computational medium for human thinking, reasoning, and aesthetic expression, her work on large-scale foundation models sits at the intersection where these expressions become efficient and precise, while remaining original and meaningfully human.
Dongyun Kim is a creative technologist and AI engineer working at the intersection of generative AI, design, and technology. His research and professional interests focus on identifying latent structures within complex phenomena and organizing them into systematic frameworks. In the context of AI, his work examines how AI models encode and interpret the world, and how real-world conditions can be represented, abstracted, and reconstructed through computational systems.
He received formal training in design and technology at Harvard Graduate School of Design and the University of Pennsylvania, supported by the Korean Government Scholarship for Overseas Study. His career experience spans architecture, Silicon Valley AI startups, and global finance, reflecting a sustained engagement with both experimental research and applied systems that inform his perspective on generative models as representational and analytical media.
