The document outlines a proposed framework for generative adversarial network-based visual interactive fashion design. It discusses using conditional GANs and SRGANs to generate high-quality fashion images conditioned on user inputs while allowing control of fine-grained styles. The objectives are to develop an interactive framework to improve image quality and finesse design capabilities. A literature review identifies gaps in existing methods, such as poor quality outputs and difficulty controlling styles. The proposed method collects and preprocesses fashion datasets before training a CGAN to generate images and an SRGAN to refine quality. Results will be evaluated and the framework aims to enhance design flexibility and image fidelity.