Unraveling the multimodal features and audience responses to thinspiration videos: A computational analysis of Chinese TikTok (Douyin)

Short-video social media platforms have emerged as powerful channels for disseminating body image-related content, particularly the thin ideal. Yet much of the existing research has focused on individual-level usage patterns rather than the platform-level mechanisms that shape content exposure. A platform-level perspective, however, is crucial for understanding how body image messages are algorithmically curated and how multimodal features influence audience responses. To address this gap, we developed a computational framework comprising body iconography, communication strategies, and audiovisual aesthetics to investigate thinspiration videos on Douyin (Chinese TikTok). Specifically, integrating computer vision, natural language processing, large language models, and audio signal processing, we extracted textual, visual, and auditory features from 757 trending thinspiration videos tagged with #thin and linked these features to nearly three million associated comments to assess large-scale audience responses. Results showed that trending thinspiration videos promoted the thin-ideal, along with the fit-ideal and beauty-ideal, often with gendered portrayals of the human body. Multimodal features such as body size, evaluative rhetoric (thin-praise, fat-stigma), and specific aesthetic patterns were significantly associated with engagement metrics (likes, comments, reshares) or the prevalence of appearance-focused comments. Gender differences were pronounced: female-focused videos featured smaller body sizes, higher beauty scores, and greater objectification, whereas male-focused videos emphasized muscularity and informational framing. This study advances understanding of body image formation in algorithmically mediated environments and offers a scalable methodological framework for multimodal content analysis. The findings offer insights for content creators, platform regulators, and policymakers to promote healthier digital body image environments.