You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Apr 18, 2022. It is now read-only.
我用你的这个方法mtts.py生成训练的lab,在标贝数据集上,如果去掉文本中的
#1、#2、#3、#4
这些韵律标签,用merlin训练能得到一个还算ok的合成结果。但是如果携带这些韵律标签去生成lab,同样的参数下,合成的效果就很差,差到听不得。请问这是什么原因呢?
是否有Prosody Mark对产生lab的区别只有:
韵律词层:
/D:d1=d2!d3@d4-d5&
以及 语句层F中的f4
The text was updated successfully, but these errors were encountered: