Cross-Media Animation Narrative Innovation in Short-Video Ecosystems: A Multimodal Generation Framework for Fashion Textile Brands

Abstract

Cross-media animation narrative is reshaping content production in short-video ecosystems, while fashion textile brands still face problems such as insufficient narrative density, fabric texture distortion, and multimodal semantic misalignment on platforms such as Douyin and Xiaohongshu. To address these challenges, this study constructs CrossNarrativeNet, a cross-media narrative framework integrating textual semantic parsing, fabric texture modelling, and temporal animation generation. A structured storyboard parsing module models narrative rhythm; a texture-aware diffusion model renders silk, cotton-linen, intelligent fibers, and other fabrics with high fidelity; and a multimodal alignment network maintains semantic consistency across text, visual, and background music channels. For intelligent textile scenarios, the framework also supports functional-material visualization, including chromic or responsive fiber behavior driven by simulated physical stimuli, which is relevant to future electromagnetic-responsive textile displays and sensor-integrated wearable media. The self-constructed FashionShort dataset covers six categories and twelve typical narrative archetypes. Experimental results show that the proposed framework outperforms existing methods in generation quality, narrative completeness, and predicted user dwell time. Cross-platform deployment on Douyin vertical screen, Xiaohongshu square screen, and WeChat Channels widescreen further validates its generalization capability. The study also proposes application guidelines from the perspectives of platform compliance, intellectual property, and designer rights, providing technical and governance references for intelligent textile brands and multimodal content propagation in short-video ecosystems.