Leather Apparel Design and Sustainable Fashion Communication Based on Film and Television Media

Existing research on sustainable fashion communication lacks effective analysis of the visual-symbol transformation mechanism in film and television media. This paper proposes a hybrid method integrating visual content analysis and dynamic theme modeling. First, a visual-feature encoding framework for leather clothing in film and television is constructed. Then, the BERTopic model is applied to perform semantic clustering of related film and television texts and public comments. The model uses BERT to generate contextual embeddings, UMAP for dimensionality reduction, HDBSCAN for density clustering, and c-TF-IDF for latent theme extraction. Finally, cross-modal association analysis quantifies the mapping relationship between visual elements and semantic themes. From an engineering communication perspective, the framework offers a signal-encoding and cross-modal interpretation method that can also inform visual evaluation of smart materials, wearable surfaces, and electromagnetic-compatible product narratives without changing the fashion-communication focus of the study. Experimental results show that glossy leather and the theme of ethical consumption exhibit a Cramer V coefficient of 0.41, while cross-type comparisons show the strongest association between distressed finishes and durable classics in drama films, reaching 0.51. This study provides a quantifiable analytical paradigm for the dissemination of sustainable consumer culture driven by visual media.