A structured Kano-AHP framework for AI-assisted generative media production: experimental evidence from short-form animation design

Although generative AI has accelerated parts of media production, the quality of final outputs still depends on how clearly requirements are defined, prioritized, and maintained across prompting, revision, and selection. This study evaluates a structured workflow that integrates the Kano model and the Analytic Hierarchy Process (AHP) into the Double Diamond process for AI-assisted media production. Twenty practitioners each produced one anime-style vertical clip under either a baseline workflow or a structured package workflow, yielding 20 final clips (10 per condition). Five experts rated all clips on narrative coherence, visual style consistency, emotional expressiveness, motion smoothness, and technical quality. In a randomized partial-viewing design, 60 audience viewers each rated eight clips, giving 24 audience evaluations per clip. Conditions were compared using descriptive statistics, Hedges’ g, intraclass correlation coefficients, and linear mixed-effects models with Holm adjustment. Across all five dimensions, the package condition received higher ratings than the baseline condition in both expert and audience evaluations. Audience effect sizes ranged from g = 0.74 to 0.99, expert effect sizes from g = 0.98 to 1.32, and expert inter-rater reliability ranged from ICC(2,k) = 0.70 to 0.86. These findings suggest that, under a common toolchain, explicit requirement structuring and criterion weighting can improve the judged quality of AI-generated media outputs. The current design, however, evaluates the structured package as a whole rather than the separate effects of Kano and AHP.