Talk:Luo Tianyi/@comment-77.111.244.21-20180501140912/@comment-53539-20180503165347

I did wonder this myself.

To me my figuration was that the original VA was used for the bulk of the voice and the second for odd samples.

I believe that most voicebanks are defined in how they sound by 40% of the voicebank and the pronunciation of words by a further 30% on top as these are all the core sounds. The last 30% is just smoothing out samples like triphones and doesn't have a great deal of impact on the voice or how it sounds. In theory, if 40% of the voicebank is using the original voice actress, the voice should mostly sound the same. You may hardly be able to tell there is a second actress in use. Still, the fact one is used means this is going to impact the voicebank in certain ways, even if small.