Self-contained analysis of whether the 6064 MOSS character reference voices form discrete clusters or a continuum, across three embeddings (ECAPA-TDNN identity, VoiceNet perceptual, VoiceCLAP audio) × {KMeans, Ward, HDBSCAN} × a k-sweep.
Live page: https://laion-ai.github.io/moss-voice-clustering-analysis/
The page (docs/index.html) is fully self-contained: inline-SVG charts (no external
libraries) computed from the metric sweep, plus an audio neighbor-cluster audit.