How can I generate without -1 and Label with accurate time information? - Githubissues

hyperzlib / Sinsy-Remix

The HMM-Based Singing Voice Syntheis System Remix "Sinsy-r"

Other

41 stars 25 forks source link

How can I generate without -1 and Label with accurate time information? #10

Closed yuuSiVo closed 4 years ago

yuuSiVo commented 5 years ago

How can I generate without -1 and Label with accurate time information?

this time information is too bad

The time information is the same for consonants. Is there a way to fix this?

hyperzlib commented 4 years ago

-1 means generate time from model.

hyperzlib commented 4 years ago

you should use speech align tools to generate the 'mono label' for train.