How do I make a Japanese CV-VC voicebank?

[Arissa]

I'm finally able to make my utauloids, and the first I want to have a Japanese CVVC bank. For one I want to know how to record the sounds, not the Japanese syllables but the parts that aren't included in regular CV, like ak and such. Any help would be appreciated

 

[RaccoonButler]

Usually for cvvc you record it in strings the way you would record VCV. (Ex. "Kakikukeko" would all be one recording.)
Which reclist are you using? It's usually better if they have transitional sounds (ex. Getting the ak sound from "kaka" instead of just ak)

But if it is just the ending sound, then you record it just as it is written

 

[Arissa]

Usually for cvvc you record it in strings the way you would record VCV. (Ex. "Kakikukeko" would all be one recording.)
Which reclist are you using? It's usually better if they have transitional sounds (ex. Getting the ak sound from "kaka" instead of just ak)

But if it is just the ending sound, then you record it just as it is written[/QUOTSorry for the late reply. I won't be able to use utau until later this month, but the reclist I was thinking about using is the one that was used for Namine Ritsu's CVVC lite, if I'm correct

[doublepost=1486246722][/doublepost]Oops, worded that wrong, my bad. Namine Ritsu's CVVC Eve voicebank

 

[kimchi-tan]

So what you're asking for are all the VC sounds required in CVVC Japanese?

 

[Arissa]

So what you're asking for are all the VC sounds required in CVVC Japanese?

So what you're asking for are all the VC sounds required in CVVC Japanese?

I'm looking for how to go about the recording process, like, I k ow how CV is recorded, but how would I record CVVC?

 

[수연 <Suyeon>]

VC transitions are created when you record in strings. Ex. かか will give you - か, a k, か. There's no special/separate process to recording VC. You simply record your strings without pausing.

recording sample: https://clyp.it/e5zgpag2

To configure the transition...

Consonant should be relatively high - 250 or 300... depends on length of the cutoff. I usually have my cutoff at -400.
Preutterance should be set where the vowel begins to fade out.
Overlap should be Preutterance ÷ 2.

For hard consonants (b, d, g, k, p, t) the cutoff should be right before the next consonant. Other consonants (f, h, j, l, m, n, r tap, rhotic r, s, sh, th, v, z) should have some retained.

 

[Arissa]

VC transitions are created when you record in strings. Ex. かか will give you - か, a k, か. There's no special/separate process to recording VC. You simply record your strings without pausing.

recording sample: https://clyp.it/e5zgpag2

To configure the transition...

Consonant should be relatively high - 250 or 300... depends on length of the cutoff. I usually have my cutoff at -400.
Preutterance should be set where the vowel begins to fade out.
Overlap should be Preutterance ÷ 2.

For hard consonants (b, d, g, k, p, t) the cutoff should be right before the next consonant. Other consonants (f, h, j, l, m, n, r tap, rhotic r, s, sh, th, v, z) should have some retained.

Oh, so if I were recording something like kaka and then Sasa and so on, I'd record kaka first, then Sasa later on in the same recording?

 

[Kiyoteru]

You said you already had a reclist, no? You can load it up in OREMO, an application for recording UTAU banks. It will split each sample in the list, and when you record, automatically save the files.

 

[수연 <Suyeon>]

You record according to how a given reclist is organized. I personally record 2 mora (similar to Namine Ritsu EVE) and in strings that have a whole set. For ex. kaka-kiki-kuku-keke-koko; kya-kyu-kyo. There's also the option to record a string as such... kakakikukekoko. CVVC reclisting structure can be customized for preference.

 

[Arissa]

You said you already had a reclist, no? You can load it up in OREMO, an application for recording UTAU banks. It will split each sample in the list, and when you record, automatically save the files.

Yes, although I had initially planned on using audacity. Which do you think would be best?

 

[Kiyoteru]

OREMO is the most efficient way to record a bank. If you were planning on doing any processing to the audio with Audacity, you can still do it in bulk after recording.

 

[Arissa]

OREMO is the most efficient way to record a bank. If you were planning on doing any processing to the audio with Audacity, you can still do it in bulk after recording.

Oh, I see. Thanks, I'll use oremo for recording instead

 

[Arissa]

OREMO is the most efficient way to record a bank. If you were planning on doing any processing to the audio with Audacity, you can still do it in bulk after recording.

So how do I use oremo? I downloaded it and my system locale is Japanese so that's taken care of, but I'm unsure of how to record a bank with it

 

[Kiyoteru]

I'll assume for now that you have windows version 3.0 b170312, the latest one. (It's actually supposed to also work on computers without Japanese locale, but I've only tested that on windows 7.)
If you already have a txt file for the recording list, open it up by going to File > Load Voice List. The list will then be opened, and it will automatically be on the first one.
If you want to extract the reclist from an existing voicebank, select that voicebank by clicking on the bottom bar or open folder icon and searching for the folder. Then go to File > Generate Voice List from wav Files in Recording Folder.
Check your audio settings with Option > Audio I/O Settings to make sure you have the correct microphone selected. Make sure you select the correct folder to save all your samples to, by clicking on the bottom bar or open folder icon in the main window and finding it. By default it's set to "result" in the OREMO folder. Press and hold the "R" key to record. When you go onto another sample, by clicking on the list or using the arrow keys, the previous one will automatically be saved.

 

[Arissa]

I'll assume for now that you have windows version 3.0 b170312, the latest one. (It's actually supposed to also work on computers without Japanese locale, but I've only tested that on windows 7.)
If you already have a txt file for the recording list, open it up by going to File > Load Voice List. The list will then be opened, and it will automatically be on the first one.
If you want to extract the reclist from an existing voicebank, select that voicebank by clicking on the bottom bar or open folder icon and searching for the folder. Then go to File > Generate Voice List from wav Files in Recording Folder.
Check your audio settings with Option > Audio I/O Settings to make sure you have the correct microphone selected. Make sure you select the correct folder to save all your samples to, by clicking on the bottom bar or open folder icon in the main window and finding it. By default it's set to "result" in the OREMO folder. Press and hold the "R" key to record. When you go onto another sample, by clicking on the list or using the arrow keys, the previous one will automatically be saved.

Oh, the one I downloaded says 3.0-b170308beta, but your steps worked. Thanks I'll start recording now

 

[Arissa]

I'll assume for now that you have windows version 3.0 b170312, the latest one. (It's actually supposed to also work on computers without Japanese locale, but I've only tested that on windows 7.)
If you already have a txt file for the recording list, open it up by going to File > Load Voice List. The list will then be opened, and it will automatically be on the first one.
If you want to extract the reclist from an existing voicebank, select that voicebank by clicking on the bottom bar or open folder icon and searching for the folder. Then go to File > Generate Voice List from wav Files in Recording Folder.
Check your audio settings with Option > Audio I/O Settings to make sure you have the correct microphone selected. Make sure you select the correct folder to save all your samples to, by clicking on the bottom bar or open folder icon in the main window and finding it. By default it's set to "result" in the OREMO folder. Press and hold the "R" key to record. When you go onto another sample, by clicking on the list or using the arrow keys, the previous one will automatically be saved.

So after I've finished, how do I apply this voicebanks to the utau program?

 

[Kiyoteru]

Oh, you're not familiar with that...? (Usually, people using CVVC already have some experience.)

In the folder where UTAU is installed, is another folder named "voice". Put the folder of your recordings in there. When you open UTAU, you can go to Project > Project Properties and select the new voicebank from the drop down list.
But before you can actually use it, you need to configure the OTO so that UTAU knows what each part of the sample is, and align the timing properly. You can access this from Tools > Voice Configurations. (I'm not sure exactly what it is, because my version of UTAU is in a different language.) You can also use the keyboard shortcut ctrl+G.
When you edit this, it saves everything to a file named "oto.ini" in the voicebank folder. If you based your voicebank exactly off another one, where all the file names are the same, you can copy and paste the oto.ini to use as a base. Or, if the reclist you're using came with a base oto, you can put that in the folder. But if there is nothing to base it on, you will have to set it up by yourself.
Though I don't currently have a tutorial on how to OTO a Japanese CVVC bank, I'm sure other people do. I do, however, have a tutorial for fine tuning an English bank, and some of the same principles will apply. http://arpasing.neocities.org/resources/vb-creation.html

CVVC can be rather complicated, and will probably be a frustrating introduction to UTAU. So, even if you don't think it sounds good, it's a good idea to make a CV voicebank just to learn basic functions of UTAU. This is my beginner's OTO tutorial for CV voicebanks. http://utaforum.net/resources/cv-otoing-guide.269/ Please work hard and don't give up!

 

[Arissa]

Oh, you're not familiar with that...? (Usually, people using CVVC already have some experience.)

In the folder where UTAU is installed, is another folder named "voice". Put the folder of your recordings in there. When you open UTAU, you can go to Project > Project Properties and select the new voicebank from the drop down list.
But before you can actually use it, you need to configure the OTO so that UTAU knows what each part of the sample is, and align the timing properly. You can access this from Tools > Voice Configurations. (I'm not sure exactly what it is, because my version of UTAU is in a different language.) You can also use the keyboard shortcut ctrl+G.
When you edit this, it saves everything to a file named "oto.ini" in the voicebank folder. If you based your voicebank exactly off another one, where all the file names are the same, you can copy and paste the oto.ini to use as a base. Or, if the reclist you're using came with a base oto, you can put that in the folder. But if there is nothing to base it on, you will have to set it up by yourself.
Though I don't currently have a tutorial on how to OTO a Japanese CVVC bank, I'm sure other people do. I do, however, have a tutorial for fine tuning an English bank, and some of the same principles will apply. http://arpasing.neocities.org/resources/vb-creation.html

CVVC can be rather complicated, and will probably be a frustrating introduction to UTAU. So, even if you don't think it sounds good, it's a good idea to make a CV voicebank just to learn basic functions of UTAU. This is my beginner's OTO tutorial for CV voicebanks. http://utaforum.net/resources/cv-otoing-guide.269/ Please work hard and don't give up!

Oh, thank you lol I actually figured that out by the time you replied ^^; But another issue that I had when trying to make Defoko sing was that even when using the right kana, here I use letters, I can't hear the voice, even after I press play

 

[Kiyoteru]

Does the same thing happen with other voicebanks, that aren't your own?

 

[Arissa]

Does the same thing happen with other voicebanks, that aren't your own?

So far I only have Defoko and the voicebank I just made. For Defoko I used hiragana but I couldn't hear her voice :/ For mine I tried playing notes but there still was no sound