Condensed clips have emerged as a novel form of user-generated content (UGC), offering a compelling and concise alternative to traditional full-length movies. While Danmaku in condensed clips enrich the co-viewing experience, they also induce moral disengagement due to anonymity, leading to aggressive behaviors in the likes of insults, obscenity, and profanity in the comments. Grounded in Cognitive Dissonance Theory (CDT), this study investigates the detrimental effects of aggressive Danmaku on the performance of condensed clips and explores the moderating role of the original work's attributes. Through an empirical analysis of a large-scale dataset, we demonstrate that a high proportion or intense concentration of aggressive Danmaku can negatively impact video performance. Furthermore, the reputation of the original work amplifies the negative effects of the proportion of aggressive Danmaku while reducing the impact of its concentration. In contrast, the nostalgic appeal of the original work intensifies the detrimental effects of both the proportion and concentration of aggressive Danmaku on video performance. Additionally, we conducted two post-hoc analyses: one to explore the existence of potential nonlinear relationships, and another to capture the diverse forms of aggressive Danmaku based on content and style. This study contributes to extant literature on cyber aggression and viewer behavior in consuming condensed clips, highlighting the dark side of Danmaku on video performance as well as yielding practical insights for UGC platforms and content creators. • This study advances understanding of viewing behavior in consuming condensed clips. • Enriches research on cyber aggression by examining harmful effects of Danmaku . • Explores cognitive dissonance from aggression in co-viewing condensed clips. • Advances understanding of Danmaku by showing its dual effects on UGC performance. • Offers insight: Original work's attributes moderate effects of aggressive Danmaku .