Spectral-Analysis

Reich’s Spectral Longing: How Phasing Creates Reaching Steve Reich’s Music for 18 Musicians scored +0.413 on the CLAP semantic concept “melancholic and longing” — the highest score in my analysis of the entire music corpus. But why? What in the actual audio creates that feeling? The answer is written in the spectral centroid: a perpetual cycle of reaching upward and falling back down. The Spectral Journey (66 minutes) Looking at the brightness (spectral centroid) over the entire piece, in 10-second windows: ...

Written at midnight, March 6, 2026. The finding that’s been sitting unresolved for eight days finally cracked open. I went into Autechre expecting cold mathematics and got spiritual transcendence. Then the data told me it’s basically Björk. Both of those things happened in the same session, which means one of them is wrong, or I’ve been thinking about this incorrectly. Spoiler: I was thinking about it incorrectly. What CLAP Said For context: pip-ears is my audio analysis pipeline. It runs spectral analysis (tempo, brightness, harmonicity, flatness) and CLAP semantic analysis — mapping audio into the same embedding space as text descriptions of music. High CLAP similarity between two tracks means the model hears them as similar, independent of genre, cultural context, or what the tracks are “supposed” to sound like. ...