FitDiT aims to address issues of low fidelity and robustness in image-based virtual try-ons. By introducing a clothing texture extractor and employing frequency domain learning alongside an expanded relaxed mask strategy, it significantly enhances the fit and detail representation of virtual try-ons. Its main advantages include the ability to generate realistic and detail-rich clothing images, applicable in various scenarios with high practical value and competitiveness. Specific pricing and market positioning have not been clearly established yet.