This article presents an affordable approach to timbral analysis of music using ordinary computers, accessible audio software, and AI-assisted coding. It argues that while pitch, tempo, and dynamics can be represented on a single scale, timbre and harmony are multidimensional and are better understood as positions in a space defined by several interacting features.
The article shows how technical descriptors such as spectral centroid, flatness, roughness, and sensory dissonance can be linked with listener-based terms such as brightness or harshness, and explored through methods such as principal component analysis and multidimensional scaling. It also shows how AI-assisted programming can help build custom tools for spectrograms, overtone isolation, and frame-by-frame feature extraction, including a web-based Timbral Trajectory tool for visualizing timbre over time.
At the same time, the article stresses that analysis measures sound signals, not experience itself. For that reason, technical methods must remain connected to listening, introspection, and listener research. AI is treated here not as an oracle, but as a practical partner in musician-centered analysis and teaching.
