Fashion brand clothing possesses unique characteristics. With the improved performance of deep learning models, there have been many reported studies in the field of fashion such as virtual try-on and attribute editing. Nevertheless, extant research has been limited in its ability to systematically capture and represent brand-specific visual characteristics. Due to the differences in environment, composition, and human posture across photographs, understanding clothing characteristics through usual image learning approaches proves challenging. In this study, we explored image generation which reflects brand-specific features for the purpose of deep learning to understand the features of clothing. With the support of a fashion brand specializing in Lolita fashion, we utilized Stable Diffusion to generate high-quality images. To obtain training-ready images, we applied SAM2 and YOLO11 to product images, removing backgrounds and facial features while retaining only the clothing required for training. Words related to color, impressions, and clothing features were extracted from the product description reflecting the brand design, and generated text prompts so that the user can generate their preference. Then, we compared the original product images and the generated images, and evaluated the quality of image generation which captured brand-specific features and the effect of the prompts. Furthermore, we explored LoRA-based model merging as an extension, investigating whether it can integrate features of multiple products to produce images of new categories.