Description

Compared to the last two models this one is the cleanest of the three- obviously because the source audio is the cleanest of the three. Unfortunately Cortana still doesn't quite have the range I was hoping for but as a speech-based model and not a singing-focused model I'd argue it's worth its salt. Model can be found here: < Included too below is the loss graph for this model trained up to the full 1000 Epochs.