Understanding AI Animation Through Principle Patterns: A Framework for Future Development

This study presents a systematic evaluation of how generative AI animations adhere to the fundamental principles of animation established by Thomas and Johnston. Through rigorous analysis of five AI-generated animations evaluated by three expert animators against ten classical principles, we identified distinct patterns in AI's animation capabilities. Our methodology combined expert evaluation with comprehensive statistical analysis, including inter-rater reliability assessment (κ = 0.84), factor analysis, and capability mapping. Results revealed three clear capability clusters: consistently implemented principles (Staging, Solid Drawing, Appeal; 100% presence), variably implemented principles (Timing, Straight-Ahead Action/Pose-to-Pose, Follow-Through; 20-73% presence), and consistently absent principles (Squash and Stretch, Anticipation, Exaggeration; 0% presence). The findings demonstrate that current AI systems excel at spatial and compositional elements but struggle with dynamic, physics-based principles. This research contributes to the field by establishing a structured framework for evaluating AI animation capabilities, providing actionable insights for both developers and practitioners, and suggesting effective hybrid workflow approaches. Our analysis indicates clear pathways for future development, emphasizing the need for enhanced temporal control and physics-based animation capabilities in AI systems.