Mean 9-feature signatures of the 24 largest genres, on identical axes
The Sonic Genome — Genre Fingerprints. Mean audio-feature signature of the 24 largest genres by usable track count (n = 997–1,000 tracks per genre; 2022 Spotify snapshot). Genre sizes are near-uniform by construction (each seed contributed ~1,000 tracks), so this selection effectively excludes the seeds with the most tempo/time-signature detector failures. All panels share one radial scale (0–1) and one fixed spoke order, so silhouettes compare directly: detroit-techno and chicago-house carry tall instrumentalness spikes absent everywhere else; cantopop and mandopop lean acoustic and low-valence; salsa is the happiest panel (mean valence 0.81); and the near-twin silhouettes of punk and punk-rock show sibling seeds sharing one fingerprint — signature as signal, not noise. Seven features are native 0–1 scores; loudness* and tempo* are min-max scaled across all 114 genre centroids (loudness -24.5 to -3.8 dB; tempo 98 to 155 BPM). Radar shape depends on spoke order (identical everywhere); enclosed area is not a meaningful quantity. The first panel carries the spoke labels (abbreviated) for the whole grid; hover any vertex for exact feature values.
Centroids: data/processed/genre_centroids.parquet (unweighted means, quality-flagged tracks excluded; lineage in genre_centroids.lineage.json). Panel title color = super-genre family (see Genre Galaxy legend). Source: Kaggle maharshipandya/-spotify-tracks-dataset; Spotify deprecated the /audio-features endpoint Nov 2024.