Description

Thanks to for 45 minutes of sample audio, and his previous versions, he inspired me to use the FULL 45 minutes, it took a while and tried to assassinate my GPU, but I got there. It is trained to 256 epochs, exactly 68,000 steps, due to it overtraining past that point. It sounds REALLY good and has a good vocal range, and is great for TTS. Give it a shot. It works best with harvest/crepe!