Ivan Barajas Vargas’ Post

Generative AI is creating exciting opportunities in most every industry, with near limitless potential in front of it. Fashion is no different – LLMs are seeing great success with use cases ranging from advanced planning to e-commerce copywriting. They are helping fashion companies manage inventory levels and consumer data but they do little to support a crucial need – choosing the right product. From their introduction, it was clear that Latent Diffusion Models have the potential to disrupt the fashion industry by doing exactly that. • Why Latent Diffusion? Simply, Latent Diffusion Models (LDMs) create images. LDMs are a class of generative deep learning models that are trained on images and text describing them. These models can be prompted to create new images based their training, which can be via big generic base models, smaller highly specialized proprietary ones, or commonly a mix of both. Inputs can include prompts and other images, and the results are high-resolution images that require relatively low computational resources to generate. • Latent Diffusion models aren’t perfect. Far from it, they need to be fitted, just like clothing and shoes. The tools are incredibly complex. There are dozens of intimidating parameters to select, some that influence the output dramatically. They are built for hobbyists to create intricate workflows and very much like 3D, their usage in enterprises would be limited to a small highly-trained team. Even then, the outputs are generic - it doesn’t understand your product or brand’s DNA. Finally, readily available tools don’t scale - most are built without the features enterprises require, and they aren’t tailored to fashion’s unique processes. • With Greige Create, we took on these challenges head on. We start with a unique approach to AI where we train smaller, fine-tuned modular models with our client’s assets. Our proprietary pipelines are built on an infrastructure with the privacy and control that enterprises demand, and our fashion specific workflow seamlessly ties together LDMs and LLMs while expecting inputs that are typical in today’s process and produce outputs of the quality and fidelity that are demanded. • Finally, we applied an important learning from our 25 years in this space – adoption drives success and usability drives adoption. We bring a simple UI where the complexity has been moved into the background. Configurable presets and tools are set once so the user is presented with few distractions and clicks. We also recognize that user experience is more than UI, and the only approach better than a few straightforward clicks is to eliminate the work altogether. Through AI-driven automation and APIs that allow for data to be shared with systems like PLM, DAM, and visual line planning we are taking our first steps towards that vision.   Greige Software – Enterprise AI for Fashion. Shorten the calendar. Widen the funnel. Reduce Operating Expense

Generative AI is creating exciting opportunities in most every industry, with near limitless potential in front of it. Fashion is no different – LLMs are seeing great success with use cases ranging from advanced planning to e-commerce copywriting. They are helping fashion companies manage inventory levels and consumer data but they do little to support a crucial need – choosing the right product. From their introduction, it was clear that Latent Diffusion Models have the potential to disrupt the fashion industry by doing exactly that. • Why Latent Diffusion? Simply, Latent Diffusion Models (LDMs) create images. LDMs are a class of generative deep learning models that are trained on images and text describing them. These models can be prompted to create new images based their training, which can be via big generic base models, smaller highly specialized proprietary ones, or commonly a mix of both. Inputs can include prompts and other images, and the results are high-resolution images that require relatively low computational resources to generate. • Latent Diffusion models aren’t perfect. Far from it, they need to be fitted, just like clothing and shoes. The tools are incredibly complex. There are dozens of intimidating parameters to select, some that influence the output dramatically. They are built for hobbyists to create intricate workflows and very much like 3D, their usage in enterprises would be limited to a small highly-trained team. Even then, the outputs are generic - it doesn’t understand your product or brand’s DNA. Finally, readily available tools don’t scale - most are built without the features enterprises require, and they aren’t tailored to fashion’s unique processes. • With Greige Create, we took on these challenges head on. We start with a unique approach to AI where we train smaller, fine-tuned modular models with our client’s assets. Our proprietary pipelines are built on an infrastructure with the privacy and control that enterprises demand, and our fashion specific workflow seamlessly ties together LDMs and LLMs while expecting inputs that are typical in today’s process and produce outputs of the quality and fidelity that are demanded. • Finally, we applied an important learning from our 25 years in this space – adoption drives success and usability drives adoption. We bring a simple UI where the complexity has been moved into the background. Configurable presets and tools are set once so the user is presented with few distractions and clicks. We also recognize that user experience is more than UI, and the only approach better than a few straightforward clicks is to eliminate the work altogether. Through AI-driven automation and APIs that allow for data to be shared with systems like PLM, DAM, and visual line planning we are taking our first steps towards that vision.   Greige Software – Enterprise AI for Fashion. Shorten the calendar. Widen the funnel. Reduce Operating Expense