Closed fakerybakery closed 11 months ago
Nope; any ideas for how such an eval would work?
Hmm, I don't know how that would work. Thanks though!
One thing - so with this eval, running it on samples generated by a non-AI TTS system (ie espeak, macOS "say" command, festival, etc) would give similar results to running it on samples generated by Tortoise/XTTS?
Hi, Does this repo also evaluate how realistic the voices are?