How do I make a Japanese CV-VC voicebank?

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
Try downloading other voicebanks and/or using premade USTs to see if those work. That way, we can see whether it's an issue with UTAU, or just Defoko.
 
  • Like
Reactions: Arissa

Arissa

Ritsu's Renegades
Defender of Defoko
Thread starter
I think it's a problem with Utau, but I'm not sure what...
 

Arissa

Ritsu's Renegades
Defender of Defoko
Thread starter
Never mind! I reinstalled utau and Defoko sings!!! I think it was because I downloaded the program before my locale was switched
[doublepost=1489903044,1489901882][/doublepost]But the problem I have now is when I listened to my utau sing. The voice itself sounds fine but I noticed (as a result of me being a recording noob) that there's a chunk of space before every note. For example, when she sings Koko, there's about three seconds of silence before an actual voice, so I'm wondering how I can edit the samples in oremo to get rid of unnecessary gaps like that.
 
  • Like
Reactions: Kiyoteru

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
Like I said, before you can actually use your CVVC voicebank properly, you must configure the OTO so that UTAU knows the timing of all the samples. There's no need to directly edit the audio files, and this can actually be risky because you might cut off important parts. I've written a previous post in this thread about the OTO, so please read it.
 
  • Like
Reactions: Arissa

Arissa

Ritsu's Renegades
Defender of Defoko
Thread starter
Like I said, before you can actually use your CVVC voicebank properly, you must configure the OTO so that UTAU knows the timing of all the samples. There's no need to directly edit the audio files, and this can actually be risky because you might cut off important parts. I've written a previous post in this thread about the OTO, so please read it.

Oh, I was wondering when I'd need to do that, thx. Rn I'm going to rerecord with a new reclist I found from a good tutorial that comes with a helpful oto
 
  • Like
Reactions: Kiyoteru

Arissa

Ritsu's Renegades
Defender of Defoko
Thread starter
Like I said, before you can actually use your CVVC voicebank properly, you must configure the OTO so that UTAU knows the timing of all the samples. There's no need to directly edit the audio files, and this can actually be risky because you might cut off important parts. I've written a previous post in this thread about the OTO, so please read it.
Okay, so I finally figure out how to oto lol kept looking in the wrong place before. So for things like offset and cutoff, for the former should I input when there's actually sound(singing) or just leave it be? I think I got cutoff properly, thanks to some earlier comments and research and whatnot. ^=^ Sorry if I'm burdening you, I appreciate the help
 

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
Because of the way that CVVC reclists are written, it can be difficult to give precise and helpful advice. They tend to include multiple samples in a single file, meaning that one audio file corresponds to more than one line of the OTO. The number of lines depends on the formatting of the reclist. Thankfully, if you're using a base OTO, it comes with all that stuff set up already. It's just a matter of moving the actual parameters into place.
The CV parts of CVVC are OTO'd in the same way as a bank that is CV alone. Again, I'd like to refer you to this guide. http://utaforum.net/resources/cv-otoing-guide.269/
In general, you have to move the offset, or left blank, to where the section you want to OTO begins.
If you have both [- cv] and [cv] samples, your bank is rentan style. This is the result of strings like "kaka.wav" "soso.wav" and so on. If you recorded it as "kakikukeko.wav" then there will only be plain [cv] samples. So let's say you recorded it "kakikukeko" and now you need to OTO the "ku" section. You'll bring the offset to around where the "ku" syllable begins.

It'll probably be much easier to do all this once you have a solid foundation in UTAU. If you truly insist on not making a CV bank first, then you should definitely be using other people's CVVC banks to see how they work. It can't hurt to think about the other side of things, the music making side, for a little while. It may even give you ideas for what you want to do with your voicebank. After all, it'll be pretty sad and somewhat pointless if you never use it after you make it...
 

Arissa

Ritsu's Renegades
Defender of Defoko
Thread starter
Because of the way that CVVC reclists are written, it can be difficult to give precise and helpful advice. They tend to include multiple samples in a single file, meaning that one audio file corresponds to more than one line of the OTO. The number of lines depends on the formatting of the reclist. Thankfully, if you're using a base OTO, it comes with all that stuff set up already. It's just a matter of moving the actual parameters into place.
The CV parts of CVVC are OTO'd in the same way as a bank that is CV alone. Again, I'd like to refer you to this guide. http://utaforum.net/resources/cv-otoing-guide.269/
In general, you have to move the offset, or left blank, to where the section you want to OTO begins.
If you have both [- cv] and [cv] samples, your bank is rentan style. This is the result of strings like "kaka.wav" "soso.wav" and so on. If you recorded it as "kakikukeko.wav" then there will only be plain [cv] samples. So let's say you recorded it "kakikukeko" and now you need to OTO the "ku" section. You'll bring the offset to around where the "ku" syllable begins.

It'll probably be much easier to do all this once you have a solid foundation in UTAU. If you truly insist on not making a CV bank first, then you should definitely be using other people's CVVC banks to see how they work. It can't hurt to think about the other side of things, the music making side, for a little while. It may even give you ideas for what you want to do with your voicebank. After all, it'll be pretty sad and somewhat pointless if you never use it after you make it...
This thread is really old,,, tfw you revisit a rookie help thread and you finally understand what people were trying to tell you....

Okay, so when otoing CVVC, do I do the thing with VCV where I duplicate samples? Like say in VCV I made kakikukeko or something with strings, I duplicate so I can oto for ka, ki, ku and so on and so forth? Or no?? (I haven't revisited anything CVVC in forever rip)
 

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
Yes, that's how it works. Say your voicebank is recorded rentan style, so you have "kaka.wav". For that one sample, you duplicate it for a total of 3 entries: [- ka] [a k] and [ka]. But if your voicebank is recorded stringed, like "kakikukekoka.wav", you need a total of ten duplicates. [ka][a k][ki][i k][ku][u k][ke][e k][ko][o k].

Though, if you're interested, I've made an OTO generator that can generate diphone based OTOs from whatever you throw at it. I've even uploaded some files that allow it to generate a Japanese CVVC oto.
https://github.com/adlez27/oto-generator
 
  • Like
Reactions: partial and Arissa

Arissa

Ritsu's Renegades
Defender of Defoko
Thread starter
Yes, that's how it works. Say your voicebank is recorded rentan style, so you have "kaka.wav". For that one sample, you duplicate it for a total of 3 entries: [- ka] [a k] and [ka]. But if your voicebank is recorded stringed, like "kakikukekoka.wav", you need a total of ten duplicates. [ka][a k][ki][i k][ku][u k][ke][e k][ko][o k].

Though, if you're interested, I've made an OTO generator that can generate diphone based OTOs from whatever you throw at it. I've even uploaded some files that allow it to generate a Japanese CVVC oto.
https://github.com/adlez27/oto-generator

Thanks!!! You've been a great help, I appreciate it! I've also taken to downloading some CVVC voicebanks (Yu Yao cough) to observe how they work, but I'll definitely check out that link!
 
  • Like
Reactions: Kiyoteru