VSQX to UST Sounds Off, CVVC Otoing Help

[ArtsyP!nk]

SECT. 1: VSQX to UST Sounds Off

So, I decided to pick a random .vsqx file (I chose Daughter of Evil) and create a ust for it. I was able to get the transition bits filled in, but when playing it back... it sounds really off key and cracked. Sorta like a singer going through puberty. Why is that so?

SECT. 2: CVVC Otoing Help

So for a while, I've been workin' on two voicebanks, Renaka's ACT II and the voicebank in question, Naniko Pikita/Asanka. It's been going pretty well, other voicebanks have very helpful (shoutout to the creator of Iwata Taiga), but I got a bit stuck in a few places.

They include:
(insert vowel here)b-
(insert vowel here)-
oを (but there was no wo in the line wha-)
(insert vowel here)j, f, h, ky, etc.
So could ya'll help me out?

 

[partial]

Timing issues are an oto problem almost 100% of the time. DM me with a download link to the vb and I'll fix it for you.

 

[Kiyoteru]

when playing it back... it sounds really off key and cracked. Sorta like a singer going through puberty. Why is that so?

Select all the notes, right click, and go to region properties. Make sure that the MOD parameter actually has the number 0 in it.

(insert vowel here)b-

I'm guessing that you mean samples like [a b-] [i b-] [u b-] [e b-] [o b-]. If there are other consonants in the voicebank like this (such as [a g-] [i g-] [u d-] [e d-] etc.) then this means that this is an alternative VC where the consonant is audible. For a Japanese CVVC voicebank the only VC samples you actually need are transitional ones, so plosives will never be heard. However, you may choose to include these alternate sounds to make English words easier. In that case, here's a generic diagram of how it should be oto'd.


The overlap is right before the vowel begins to fade out, so that you're blending the consistent section of the vowel with the previous note. The cutoff is further away so that the consonant is still included.

(insert vowel here)-

An ending vowel is pretty similar. You place the overlap before the vowel fades out, the preutt right when the vowel ends, and everything after should be silence.

oを (but there was no wo in the line wha-)

This looks like a VCV sample. Not sure how it got into your voicebank unless you copied and pasted someone else's OTO without double checking it was CVVC only.

(insert vowel here)j, f, h, ky, etc.

F and H are pretty much the same in terms of otoing technique. KY is entirely different. When you wrote "ETC" I was confused, because those consonants you listed don't really belong to a specific group other than "all types of consonants". In any case, I'll assume you're confused about F/H specifically, and y-blends in general.

F and H can be handled in pretty much the same way as S/Z/SH. You want to keep the entire consonant in the sample, but you don't want it to get stretched or shrunk when you're using the note.

Overlap before the vowel fades out. Preutterance where the vowel ends and consonant begins. Pink region is covering 99% of the space between the offset and cutoff. The reason we're leaving a tiny white gap is to prevent UTAU from freaking out when there's nothing given for it to stretch.

For KY/GY/NY/HY/etc. you will OTO the VC taking only the first consonant into account. Completely ignore the Y. So OTO [a ky] as if you're otoing [a k], OTO [e ny] as if you're otoing [e n], and so on.