v1 - original singing model
v2 - conversational model can’t sing (no pitch)
v3 - improved singing, with .index file

Arisu model created by @minotaurox0832(YouTube)