How to OTO growl samples?

MastermindRitsu

Momo's Minion
Hello, I hope I'm not bothering, but I'm working on my own English voicebank for my UTAU and I've been wanting to implement real growl samples into my utau's voicebank. I just listened to a bunch of heavy metal bands while OTOing, and I thought, "Hooty hoo, I want this waste of computer space to scream as beautiful as they can." So I spent like an entire day recording an entire folder of growls.

As you all know UTAU doesn't really like growling or screaming samples and doesn't pick them up, so I was wondering how I could go about OTOing the samples to show up and sound as non-distorted as possible? I know that I could use the growl resampler, or moresampler's growl feature, or the ol' vibrato growl trick, but I wanted to know if putting in real growl sounds is possible. If it isn't, that's also okay!! Any help is very much appreciated!!!

https://clyp.it/0xi0wmoe?token=e09c4aebebf3140e48c3ff3b4611d626 Heres a sample of what I mean.

Thank you for your time.
 

Awaclus

Ruko's Ruffians
Defender of Defoko
One of the things you have to do is to prevent UTAU from evening out the pitch, which you can do by editing the pitch data to make UTAU think it's already even.
 

MastermindRitsu

Momo's Minion
Thread starter
One of the things you have to do is to prevent UTAU from evening out the pitch, which you can do by editing the pitch data to make UTAU think it's already even.

How would I go about doing that? I'm sorry, I'm very stupid when it comes to stuff like this in utau :,,,,>

https://imgur.com/OlLJh8p
OlLJh8p


Do you mean I have to go in and edit the FRQ file orange line?
 

Zany

Teto's Territory
Defender of Defoko
Just oto them like you normally would. Consonants will remain as consonants even when they are growl samples.

The problem with growl samples, as implicitly stated by Awacius, is the frequency. We can't actually allot a frequency value properly because we are not sure whether growl samples are voiced or unvoiced

This means there are two things you can do, either
1) Allot a constant value of what you think the pitch of the growl sample can be, or
2) Allot 0 for samples which sounds like a growl

Personally I am on the fence with this, but I'll recommend option 2 for you because you said you aren't good with these sort of things.

To allot zero (empty value) as the frequency, the fastest way is to highlight the "to be edited" values between empty values and press avg.

https://imgur.com/nvbWZEx

If you are still confused on how to edit frq values, here's a guide
https://utaforum.net/resources/frq-file-editing-guide.8/

Oto editing wise, assuming the alias to be [a え], the end result (roughly) should be the image below. Just ensure the preutterance (red line) is directly between the non-growl and growl samples.

https://imgur.com/YzPKsy1

If you still are scared of removing the wrong frequency values, or confused at where to put the preutterance, my advice is to record the growl samples separately from the non-growl ones. But then again I guess you shouldn't waste that entire day worth of growl recordings.

Actually before you attempt these types of things, I believe you should always have a reference.
From the top of my head, a growly VB I can think of is Hanabi's Crecia THORN VB so feel free to check that out.


Last advice is that the recommended resampler for these types of edited growl samples is fresamp (especially since you don't want that time you spent editing .frq files to go to waste). Resampler.exe is unreliable and other resamplers use their own frequency files.

Or you could always try the resampler I edited, tn_fnds_PB (shameless plug) :D
https://utaforum.net/threads/tn_fndsx-resampler-download.19373/#post-138309

It uses a frequency correction algorithm which may work.
 

MastermindRitsu

Momo's Minion
Thread starter
Just oto them like you normally would. Consonants will remain as consonants even when they are growl samples.

The problem with growl samples, as implicitly stated by Awacius, is the frequency. We can't actually allot a frequency value properly because we are not sure whether growl samples are voiced or unvoiced

This means there are two things you can do, either
1) Allot a constant value of what you think the pitch of the growl sample can be, or
2) Allot 0 for samples which sounds like a growl

Personally I am on the fence with this, but I'll recommend option 2 for you because you said you aren't good with these sort of things.

To allot zero (empty value) as the frequency, the fastest way is to highlight the "to be edited" values between empty values and press avg.

https://imgur.com/nvbWZEx

If you are still confused on how to edit frq values, here's a guide
https://utaforum.net/resources/frq-file-editing-guide.8/

Oto editing wise, assuming the alias to be [a え], the end result (roughly) should be the image below. Just ensure the preutterance (red line) is directly between the non-growl and growl samples.

https://imgur.com/YzPKsy1

If you still are scared of removing the wrong frequency values, or confused at where to put the preutterance, my advice is to record the growl samples separately from the non-growl ones. But then again I guess you shouldn't waste that entire day worth of growl recordings.

Actually before you attempt these types of things, I believe you should always have a reference.
From the top of my head, a growly VB I can think of is Hanabi's Crecia THORN VB so feel free to check that out.


Last advice is that the recommended resampler for these types of edited growl samples is fresamp (especially since you don't want that time you spent editing .frq files to go to waste). Resampler.exe is unreliable and other resamplers use their own frequency files.

Or you could always try the resampler I edited, tn_fnds_PB (shameless plug) :D
https://utaforum.net/threads/tn_fndsx-resampler-download.19373/#post-138309

It uses a frequency correction algorithm which may work.


THANK YOU SO MUCH THIS IS PERFECT!!! It's exactly what I needed man. Thank you both! I'll do my best!!
 
Similar threads

Similar threads