yl4579/StyleTTS2

how to generate text sequence ?

11721206 opened this issue · 4 comments

image
wo want to know how to generate text sequence like this ?and why change the text type not like previous phone sequences? is there advantage for this kind type text ? thanks

image
wo want to know how to generate text sequence like this ?and why change the text type not like previous phone sequences? is there advantage for this kind type text ? thanks

That text is something called IPA or International Phonetic Alphabet. It is the sounds of the words and letters. We use this text to help the model understand pronunciation. You can translate/generate this text with something called a phonemizer. There are several different repos/applications of this that you can use. There is also one in my fine tuning guide which is mentioned in the readme.

lxe commented

This is done using espeak-ng phonemizer

This is done using espeak-ng phonemizer

is input pinyin sequence or phoneme sequence ?

yl4579 commented

@11721206 it doesn't work for Chinese. See yl4579/StyleTTS#10