My new Utauloid Creation [Need Help]

[Auricular]

Hey. I am making an Utau voice with a very, VERY interesting take on singing. It's a male one to start. I had someone who was gonna help me with the voice bank, but her schedule was hectic, but then she helped out by giving me samples that can make the project possible. The problem is, I need to get the Utau translated or able to sing in the language designated so that it can work. Problem is, this is my first time using the software. I was able to get it installed on my Windows 7 Acer and processing the files into usable .wav, but I don't know how to change it's language input, I guess... Anybody give me advice or anything on what I should do to get this project off the ground? This is the only place I found that I thought could help me.

I was thinking about going to VocaloidOtaku, but AzureFlux permanently closed the forum down for good and now I came here for answers.



... Also hello, I am new here.

 

[the police]

Hello! Welcome to UtaForum, nice to meet you! Your UTAU sounds interesting, I hope we'll get to see how the voice sounds soon!

Are you having trouble with the UTAU software or the UTAU voicebank? It seems you mean the latter, but I just want to double check.
The languages your UTAU can sing in depends on what's recorded and how it's recorded. Are the samples in Japanese? If so, is it in kana あ or romanji ? Or do you mean how to make it sing in the program?

I think it'd be helpful for you to come join St. Defoko's School of UTAU! It's a Discord server for UTAU users to learn or help others in a learning environment. You can also directly message a teacher for one-on-one learning sessions. You can either click or paste https://discord.gg/rSzZD9P this link into your browser. It'd also be easier to help you through there if you were to join.

 

[Auricular]

Well, I think it's software. I need to get it to sing in a language never sung before, so this is probably something a class won't help. It's probably better if someone's willing to share a few files, among other things. It's not something "conventional" to say the least either, but I thought it would work. Is there anyone who can help?

When I saw your username at first, I got nervous for some reason... Don't know why.

 

[na4a4a]

Utau is the an interface that plays sections of a voicebank. It's up to you (the voicebank maker) to add support for that language to the voicebank itself.
Utau uses a system of aliases to load the various wav files so you'll need to make sure you use those.

Since you didn't change your locale you'll now need to uninstall utau and delete the installer.
Follow the instructions here to change your locale http://www.fujitsu.com/global/suppo.../atlas/faq/setting-locale.html#procedure_win7
Then re-download a completely new installer.

 

[Kiyoteru]

What language are you looking to synthesize in UTAU? You may find that other people are already working on the same thing.
If you're working on a language that hasn't been done in UTAU before, you'll need to write a new list of samples to record. This list needs to contain all possible phonemes in the language, in a reasonable set of combinations. You can find some information to help you get started by searching for "(language) phonology" and "(language) phonotactics".

 

[Auricular]

What language are you looking to synthesize in UTAU? You may find that other people are already working on the same thing.
If you're working on a language that hasn't been done in UTAU before, you'll need to write a new list of samples to record. This list needs to contain all possible phonemes in the language, in a reasonable set of combinations. You can find some information to help you get started by searching for "(language) phonology" and "(language) phonotactics".

Say Wha?

What Language are you trying to Synthesize?

Alright, alright. Listen, I am doing an experiment, but I'm going to be honest okay... Also, I am not uninstalling UTAU, changing locale and re-installing it, Na4a4a.

I actually got some samples and also, I am not trying to synthesize Japanese, English, French, Chinese or Korean. I'm working on a Utau that can sing in Klingon and I already got samples for it. That's the truth. I already got the language table, vocal samples for it and other stuff. I just don't know how to compile it all together so that it works.

 

[Kiyoteru]

Alright, alright. Listen, I am doing an experiment, but I'm going to be honest okay... Also, I am not uninstalling UTAU, changing locale and re-installing it, Na4a4a.

You will run into perfectly preventable errors if you don't change your locale. Due to being such old software, UTAU may not work properly without it.

I actually got some samples and also, I am not trying to synthesize Japanese, English, French, Chinese or Korean. I'm working on a Utau that can sing in Klingon and I already got samples for it. That's the truth. I already got the language table, vocal samples for it and other stuff. I just don't know how to compile it all together so that it works.

Brilliant! I've done work on Klingon too! https://docs.google.com/document/d/1FQyAT_GtvxpfLBXXkDvdh3wqeD6FMRFbPPKnBSrrJok/edit
This is a basic CVVC-style recording list, that doesn't have any consonant clusters. I'm not very familiar with the language, so this is like a version 0.1 list.

Though, since I don't know who your friend is, chances are that they didn't record samples according to this. The question you seem to be asking is how to get the samples in UTAU, configure the audio timing, and then use the samples together to create a singing vocal.

 

[Auricular]

You will run into perfectly preventable errors if you don't change your locale. Due to being such old software, UTAU may not work properly without it.

I thought that's what AppLocale is for!?!

Brilliant! I've done work on Klingon too! https://docs.google.com/document/d/1FQyAT_GtvxpfLBXXkDvdh3wqeD6FMRFbPPKnBSrrJok/edit
This is a basic CVVC-style recording list, that doesn't have any consonant clusters. I'm not very familiar with the language, so this is like a version 0.1 list.

Though, since I don't know who your friend is, chances are that they didn't record samples according to this. The question you seem to be asking is how to get the samples in UTAU, configure the audio timing, and then use the samples together to create a singing vocal.

Yeah, It's Ris--
Wait, You've worked on Klingon too? Huh, who knew?
[doublepost=1522450342][/doublepost]If you said I'm wondering about configuring the audio timing and then get vocal together, then yeah I guess. I'm not sure, like I said, I've never done this before. I had an even crazier idea of proposing this to Alter/Ego, but they stopped development altogether. Plus, I'm not sure how they work.

originally, the vocals were gonna be female, but this was the best I could do on short notice. I could send you the audio files (fair warning: they're mp3. I'm still in the process of converting them in a specialized audio converter.)

Also, you already got a CVVC-style recording list? Well, this might help. Visit this link and they might help, I think.
https://www.kli.org/about-klingon/klingon-phrases/

 

[Kiyoteru]

I thought that's what AppLocale is for!?!

UTAU isn't a single all-in-one packaged EXE, there's multiple files in the folder that work together. Applocale will only simulate the correct locale environment for one element of it, so the rest will still have corrupted text. For example, check the "uta" folder in "voice". Chances are, the filenames there are completely garbled up.

Wait, You've worked on Klingon too? Huh, who knew?

I did a little search and found a page that'd be useful for further developing the list: http://www.kli.org/tlhIngan-Hol/2006/February/msg00002.html
It seems that there aren't really clusters of more than one consonant at the onset or coda alone, so the basic list may already be sufficient. But I'm interested in what recordings you already have, and how it compares to mine. How are the filenames written? IIRC my list isn't the standard way of transcribing Klingon phonemes, because I wanted to make them easier to type quickly.

 

[Auricular]

Yeah okay, Like when I tried to sample Defoko, it didn't work and for some reason, I still get an error. I don't wanna sound like a whiner, but do I really have to uninstall everything?

Also P.S. I sent you an e-mail to your inbox with the samples I talked about, in case you're wondering.

 

[Kiyoteru]

It's better to have everything set up correctly now when you're just starting, than if you waited much longer and had more things you need to reinstall.

 

[Kiyoteru]

I've received the samples you've sent me and there are quite a lot of problems that make them near-impossible to get good results with in UTAU. These are generally resolvable by the voicer who recorded them, but I want to bring them to light because it's useful information and because some of these issues are beyond simple technicalities.

In this day and age I have NO CLUE how someone managed to record audio with a sample rate of only 8kHz. I would recommend using a USB microphone. Files for UTAU need to be mono channel wav files with a sample rate of 44.1kHz and bit depth of 16, which are the typical default settings in most software. Voxengo r8brain can be used to batch-convert the sample rate.

The filenames initially lead me to believe that each file would be the single phoneme by itself, but it turned out to be something more like "nanan". However, all of the samples were quickly spoken. UTAU is meant for singing vocals, so the samples must be sung, and sustained for a longer period of time. This gives more information to the length stretching algorithms of UTAU's resamplers, so that there will be less artifacts and distortion in the resulting vocal.

Single phonemes will not blend together and transition smoothly. Since Defoko isn't working properly, take a look at another existing CV Japanese voicebank. Rather than samples like "k" "a" "i" "u", there are separate "ka" "ki" "ku" samples. This is what my previously linked list had: combinations of each consonant with each vowel. Klingon's phonotactics, as noted in the other link I shared, have a maximum syllable structure of consonant-vowel-consonant. UTAU cannot support full CVC syllables and stretch the center vowel as expected, because it can only stretch the final phoneme of a given sample. Therefore syllables are broken down into their initial and final components. The final components are VC rather than the consonant by itself, because the vowel allows it to blend into the syllable as a whole.

Simply put, this isn't usable at all. If I find some free time, I will rewrite my list according to standard transcription of Klingon, add some other samples I think would be useful, and perhaps record the samples myself strictly for testing purposes. In the meantime, you would do well to learn about how existing UTAU voicebanks are structured, and how to apply the same old concepts to a new language. For example, English voicebanks also have to handle consonants at the end of syllables. I would be willing to explain all of these necessary concepts, but you are also welcome to join my Discord server where dozens of other people are also discussing UTAU. https://discord.gg/rSzZD9P

 

[Auricular]

I've received the samples you've sent me and there are quite a lot of problems that make them near-impossible to get good results with in UTAU. These are generally resolvable by the voicer who recorded them, but I want to bring them to light because it's useful information and because some of these issues are beyond simple technicalities.

In this day and age I have NO CLUE how someone managed to record audio with a sample rate of only 8kHz. I would recommend using a USB microphone. Files for UTAU need to be mono channel wav files with a sample rate of 44.1kHz and bit depth of 16, which are the typical default settings in most software. Voxengo r8brain can be used to batch-convert the sample rate.

The filenames initially lead me to believe that each file would be the single phoneme by itself, but it turned out to be something more like "nanan". However, all of the samples were quickly spoken. UTAU is meant for singing vocals, so the samples must be sung, and sustained for a longer period of time. This gives more information to the length stretching algorithms of UTAU's resamplers, so that there will be less artifacts and distortion in the resulting vocal.

Single phonemes will not blend together and transition smoothly. Since Defoko isn't working properly, take a look at another existing CV Japanese voicebank. Rather than samples like "k" "a" "i" "u", there are separate "ka" "ki" "ku" samples. This is what my previously linked list had: combinations of each consonant with each vowel. Klingon's phonotactics, as noted in the other link I shared, have a maximum syllable structure of consonant-vowel-consonant. UTAU cannot support full CVC syllables and stretch the center vowel as expected, because it can only stretch the final phoneme of a given sample. Therefore syllables are broken down into their initial and final components. The final components are VC rather than the consonant by itself, because the vowel allows it to blend into the syllable as a whole.

Simply put, this isn't usable at all. If I find some free time, I will rewrite my list according to standard transcription of Klingon, add some other samples I think would be useful, and perhaps record the samples myself strictly for testing purposes. In the meantime, you would do well to learn about how existing UTAU voicebanks are structured, and how to apply the same old concepts to a new language. For example, English voicebanks also have to handle consonants at the end of syllables. I would be willing to explain all of these necessary concepts, but you are also welcome to join my Discord server where dozens of other people are also discussing UTAU. https://discord.gg/rSzZD9P

I get it. I agree I need to get some new samples, but I did not know that the samples were that unusable. WOW! I guess I should've known a bit better. Anyway, since they're unusable, I'm not sure how it'll go, but my partner (i.e. my mother) is gonna be the one voicing a female klingon utau I already planned out for her (UTAU is named Saj'Qa). I felt like asking Crusher-P for advice, but I felt like I bothered her enough as it is, although I'm not sure how much it'll help anyway. I'll try to see if I can dissect some other UTAUs and see if they can tell me something.

She really liked the plan after I introduced it to her and considered getting along, when she's ready, I can get her to make samples for me when she gets a chance. It'll be a while since her scheduling is sporadic and spontaneous. (You know what? That probably meant the same thing.)

Also, I never used discord before and never had to install it, but if it's that important for development purposes, I guess I'll try to log in.

 

[Kiyoteru]

I felt like asking Crusher-P for advice

Crusher-P is a music producer who occasionally used UTAU back in its early days, so even if you did get any advice it wouldn't be very useful. You made the right decision coming to UTAforum where much more experienced users can help you- especially since you ran into the one other person who has done Klingon before!

when she's ready, I can get her to make samples for me when she gets a chance. It'll be a while since her scheduling is sporadic and spontaneous.

I'd like some time to finish rewriting the recording list first.

I'll try to see if I can dissect some other UTAU and see if they can tell me something.

Dissection of the files alone won't really grant enough knowledge, so I would recommend also actually using the voicebanks in UTAU to make songs and getting a sense of how they actually work in practice.

Also, I never used discord before and never had to install it, but if it's that important for development purposes, I guess I'll try to log in.

Discord has a web client, so you can use it without needing to download any additional software.

 

[Auricular]

Pardon me, I haven't said anything in a while, but I was a bit busy actually with real life issues... So, how is the project going?

 

[Kiyoteru]

Pardon me, I haven't said anything in a while, but I was a bit busy actually with real life issues... So, how is the project going?

Sorry, it's been on the back burner. I'm currently working on a voicebank terms of use generator and another English reclist.

 

[Kiyoteru]

I've written up the new reclist, as well as a base OTO. I had a friend record a test voicebank, and this is how the result turned out: https://clyp.it/gklffpyy

Here's the reclist, OTO, and test UST: https://cloud.vocalsynth.io/3db5735cb1a550da
Let me know if you need any help with using it.