This paper explores representing creative content (lyrics, poetry, text) as multi-scale latent trajectories rather than single embeddings, and defining similarity via alignment instead of cosine distance. The goal is to capture how meaning and affect evolve over time, not just what content averages to.