This paper introduces FashionReGen, an LLM-powered system for generating fashion reports. The system, named GPT-FAR, leverages advanced Large Language Models (LLMs) to perform automatic fashion analysis and report generation. The system is designed to analyze catwalk images, extract garment information, and generate comprehensive fashion reports. The process involves three main stages: catwalk understanding, collective analysis, and report generation. In the catwalk understanding stage, GPT-4V is used to tag garments in catwalk images. This involves classifying garments into categories and generating detailed tags for each garment based on style, silhouette, neckline, length, print, pattern, detail, embellishment, and fabric. A two-stage tag cleaning strategy is then applied to ensure consistency and synonym recognition in the tags. In the collective analysis stage, the system performs statistical analysis on the tagged data. It calculates metrics such as Mix, Year-on-Year index (YoY), and evolving trend T. These metrics help in understanding the trends and changes in fashion over time. In the report generation stage, the system generates textual analysis based on the statistical data and visual elements such as charts and images. The report includes statistical charts, catwalk images, and textual analysis to provide a comprehensive overview of the fashion trends. The system is evaluated through a case study, where the effectiveness of the GPT-FAR system in generating fashion reports is demonstrated. The results show that the system can generate high-quality, comprehensive, and insightful fashion reports. The system provides a platform for automatic fashion analysis and report generation, which has significant research and application value. The paper concludes that the proposed GPT-FAR system is a novel approach to fashion report generation, leveraging LLMs to perform complex tasks in the fashion domain. The system has the potential for further enhancement, including the inclusion of more data sources and the expansion of the types of fashion reports. The evaluation of FashionReGen is another important direction for future research.This paper introduces FashionReGen, an LLM-powered system for generating fashion reports. The system, named GPT-FAR, leverages advanced Large Language Models (LLMs) to perform automatic fashion analysis and report generation. The system is designed to analyze catwalk images, extract garment information, and generate comprehensive fashion reports. The process involves three main stages: catwalk understanding, collective analysis, and report generation. In the catwalk understanding stage, GPT-4V is used to tag garments in catwalk images. This involves classifying garments into categories and generating detailed tags for each garment based on style, silhouette, neckline, length, print, pattern, detail, embellishment, and fabric. A two-stage tag cleaning strategy is then applied to ensure consistency and synonym recognition in the tags. In the collective analysis stage, the system performs statistical analysis on the tagged data. It calculates metrics such as Mix, Year-on-Year index (YoY), and evolving trend T. These metrics help in understanding the trends and changes in fashion over time. In the report generation stage, the system generates textual analysis based on the statistical data and visual elements such as charts and images. The report includes statistical charts, catwalk images, and textual analysis to provide a comprehensive overview of the fashion trends. The system is evaluated through a case study, where the effectiveness of the GPT-FAR system in generating fashion reports is demonstrated. The results show that the system can generate high-quality, comprehensive, and insightful fashion reports. The system provides a platform for automatic fashion analysis and report generation, which has significant research and application value. The paper concludes that the proposed GPT-FAR system is a novel approach to fashion report generation, leveraging LLMs to perform complex tasks in the fashion domain. The system has the potential for further enhancement, including the inclusion of more data sources and the expansion of the types of fashion reports. The evaluation of FashionReGen is another important direction for future research.