This study explores how the emotional attributes of films, inferred from online review texts, influence their diffusion patterns in the Korean movie market. Using deep learning–based emotion classification (KcELECTRA) on 182,987 audience reviews, 39 emotion variables were extracted and reduced via factor analysis to nine dimensions. The estimated Bass model parameters―innovation (p), imitation (q), and market potential (m)―served as dependent variables in regression analyses. Drawing on Affect-as-Information and Arousal-driven Sharing theories, this study elucidates the psychological mechanisms linking specific emotions to market behaviors. Results show that Tension and Irritation exert opposite effects on innovation and market potential, while Tension also enhances imitation, suggesting that films evoking suspense or unpredictability attract word-of-mouth–driven audiences and achieve stronger long-term performance. Critically, these emotional effects remain significant even after controlling for structural factors such as genre and screen availability. These findings provide new empirical evidence linking emotion-based textual signals to film diffusion dynamics and offer practical insights for producers and marketers to balance emotional tone and audience engagement strategies in the evolving, post-pandemic film industry.