A Principle-Based Evaluation of Generative AI Animation

Generative AI technologies have revolutionized animation production, yet their adherence to the foundational principles of animation remains underexplored. This study evaluates the extent to which AI-generated animations align with the ten fundamental principles defined by Thomas and Johnston. Five animations were systematically analyzed through expert evaluations and statistical methods, including inter-rater reliability, chi-square tests, factor analysis, analysis of variance (ANOVA), and logistic regression. The results revealed that while principles such as Staging, Solid Drawing, and Appeal were consistently adhered to, dynamic motion-related principles like Squash and Stretch, Anticipation, and Exaggeration were entirely absent. For intermediate principles like Timing and Follow-Through were variable, it shows the constraints of AI's ability to reproduce temporal motion control. This research contributes to the animation field by providing an analysis of the animation principles in AI-generated animation, identifying both strengths and weaknesses. Practical significance highlights the potential for hybrid workflows wherein AI is employed for static composition tasks, and nuanced motion design is reserved for human animators. The results enable the development of generative artificial intelligence technologies, and the necessity for models capable of grasping the artistry and sophistication involved in conventional animation is now clear. The study presents actionable guidance to researchers, developers, and practitioners interested in closing the gap between technological advancement and artistic design.