Tortoise should fit your use case quite well. If you want determinism then you can keep a track of reference clips, otherwise, just pass random to the voice param. https://
github.com/neonbjb/tortoi
se-tts/tree/main
…

→ View original post on X — @reach_vb