Description

Trained with 7 minutes of data. I probably could've made a better dataset by splicing different words together but that would've taken ages. The model is pretty limited since the thing was trained off a monotone tts voice that's super compressed and has a delay filter applied to everything. I would not recommend using this for singing. I also recommend turning down the pitch in the 6-12 range. You're probably gonna need to mess around with this model a lot if you want to get anything good out of it.