New preprint out! Social media is trending towards short-form video and multimodal content, but most empirical work still analyzes text, visuals, and audio separately. In this paper, we introduce a scalable + interpretable pipeline to analyze them jointly. Paper: arxiv.org/abs/2601.15278