Closed violet-duck closed 4 years ago
Yes, I have illustrated it https://github.com/begeekmyfriend/tacotron2/issues/4#issuecomment-564924780
Got it. So do I have to use --gta or I can use mel and quant dirs instand? Is there any difference between using gta and mel for wavernn training?
GTA is the evaluation from training data while mel (ground truth) is the true training data. I recommand you using GTA.
Thank you for your explanation
Hi, @begeekmyfriend I would like to know how did you do the preprocessing for wavernn since I didn't find the preprocess scripts in the repo. Also, to my understanding, this taco2 is compatible with your wavernn as I set vocoder=wavernn, is that correct? thxxxxxxx