Description

This dataset is 20 minutes of talking/singing I prefer the singing, which is optimal at 200 EPOCHS, rather than 300 EPOCHS *might retrain for singing voice* Also, artifacts are present, (greatly?) but are eliminated well with harvest or UVR