The same 11 input texts rendered by three successive fine-tunes of
openbmb/VoxCPM2. Three audio columns, nothing else, so the difference is
audible side by side in the viewer.
Column names carry the two variables that differ: corpus size and how many
Indic terms the Devanagari map covered.
column
checkpoint
corpus
Devanagari map
v1_958clips_deva37
ft_lr0.0001_10ep_deva
958 clips / ~5.1 h
37 terms
