IMAGDressing is an interactive modular clothing generation model designed to provide flexible and controllable customization services for virtual fitting systems. The model integrates the semantic features of CLIP and the texture features of VAE using a mixed attention module and integrates them into the denoising UNet, ensuring that users can control the editing. In addition, IMAGDressing also provides the IGPair dataset containing over 300,000 pairs of clothing and wearing images, establishing a standard data assembly process. This model can be used in conjunction with ControlNet, IP-Adapter, T2I-Adapter, and AnimateDiff, enhancing diversity and controllability.