Abstract: The rich content, complex texture, and vibrant colors of print patterns are fundamental to fabric production in the textile industry. Style transfer for print patterns can not only enhance designers’ creative efficiency, but also offer clients a diverse range of color options for the same pattern, thereby increasing the commercial value of the print designs. To address the problems of artifacts, lack of hierarchy, and loss of details in existing style transfer models for print patterns, we propose a style transfer model for print patterns based on hybrid structure coding and dynamic feature fusion HSC-DFF. Firstly, a hybrid encoder combining convolutional layers and Transformers is designed for rapid feature extraction from print patterns and style images, utilizing a split attention mechanism to integrate multi-scale information; then the attention layer with conditional position encoding is introduced to dynamically merge content and style features, effectively balancing semantic information with style patterns; and finally, a contrastive consistency-preserving loss function is employed to enhance the decoder’s ability to recover complete content structures and texture details by optimizing the feature differences between content and generated images. Experiments are conducted on the arbitrary image style transfer dataset composed of MSCOCO, WikiArt, and the MiHui print patterns dataset, and the results show that compared to existing style transfer models such as S2WAT, StyTr², ArtFlow, etc., HSC-DFF model achieves an average improvement of 20.2 percentage points in SSIM, with reductions of 1.5 percentage points and 8.7 percentage points in MSE and LPIPS, respectively, which can effectively preserve content structure, integrate style patterns, and maintain local details.