Making a Vocaloid Yell

[mobius017]

Any suggestions for making a Vocaloid yell? Especially for short phrases (e.g., "Stop!" "Run!" "Hide!").

Dyn and Volume will have to be up, obviously, though I'm thinking have one/both rise a little and then crash to 0 at the end of the word.

Ope is already maxed open by default, so that'll be good as-is.

Some Growl probably, for a bit of grit?

Maybe raise Gen for a bit of a husker sound?

Maybe turn up Bri (or Cle?) a little toward the middle and down toward the end?

I'm also thinking the vowels probably tend to be longer, at least for single words like these examples. Vowels probably tend to get accentuated while singing anyway, but it might be necessary to split the words into multiple notes in order to control the timing better?

No vibrato here. Remove that if it gets added, though for shorter words it shouldn't.

Some EQ/effects might help, too.

Thoughts on the above/other thoughts?

 

[lIlI]

I've found the biggest factor is pitch: I usually draw a big scoop, rising very dramatically at the end of the word. If it's possible to look at the pitch data of a real person yelling in a program like Melodyne, that would provide a great reference.

 

[mobius017]

That's a great idea! Maybe I can record myself or something and see what that looks like.

 

[mobius017]

The suggestion to look at examples in Melodyne and work with pitch bends are really good advice! I'm also finding that Miku's Solid VB and E.V.E.C. are still really helpful. Adding some XSY with her Dark VB gives her voice some more body, as well.

I get a weird sort of clipped sound when I use E.V.E.C.'s voice release, though. Anyone else get that? Is it resolvable? It seems like I can lessen it by choosing a short breath rather than a long one, but it's still there. It's strongest after [a].

Edit: Still interested in people's E.V.E.C. experiences, but I think I found a workaround: rather than relying on voice release to extend notes, add another short note, 1 semitone higher, with the [-] phoneme afterwards. This keeps the pronunciation strong (rather than diminishing at the end), and since I'm turning the DYN and Volume down at the end of a yell ATM, you don't hear it anyway.

 

[mobius017]

Update: Here's what it sounds like currently. I'm relatively pleased with what I've achieved (given that I don't tune much). I feel like it is reaching toward what I have in mind, but it's not quite there yet. I'm going for more of a martial arts vibe (something like the loud "Ha!" that MikitoP achieves at the very end of "Neppuu"), but she doesn't sound aggressive enough, and I'm musing on what I can do to make that happen....

Thoughts anyone?

I feel like the issue is somewhere between the volume (need to make it louder) and the tapering at the end of the notes (in PIT, VOL, and DYN). I don't *think* raw volume alone will solve it--a loud kitten is still a kitten, not a saber-toothed tiger. And there needs to be some kind of tapering at the ends, but it might need adjustment somehow.

(Drat, the render didn't catch the first shout from the screenshot. So you're hearing shouts 2-4.)

 

[mobius017]

Hm, I think these sound a bit better. Especially if I jack up the volume in the DAW, and add a good bit of reverb and maybe some delay?

 

[AddictiveCUL (Add)]

Hm, I think these sound a bit better. Especially if I jack up the volume in the DAW, and add a good bit of reverb and maybe some delay?

I will try to help you later, now is too late here (02:47 AM XD)

 

[IO+]

Don't be afraid to push it harder, it might sound aggressive when listening in solo but in the context it is may not sound aggressive enough especially in the dense of content, feel free to YEET that growl/brightness parameter to the max just short amount of time (like 200-300ms) and maybe put a small amount of clearness on top of it


In the mix i usually export or cut the growling as separate part.
It's look like this: Track1: vocal(no growl) Track2: vocal(growl max) and export it or you can do it normal way, cut and duplicate the growl part.

When processing, you might try to clip it compress it make them look like a juicy grill fat sausage, after that you raise the gain on growl part up until you happy.

 

[AddictiveCUL (Add)]

IO+ totally slayed, but I will try to help you with whatever I found. First things first, I totally agree with IO when he says that u shouldn't worry that much about the solo vocal. Yeah, sure there are a lot of things to/you can do, but it's like he says: what sounds in solo can be different of how it works in context. So, for you who wants a scream in a specific part a recommend you to do like IO said. Try it in a different track and see how it's going in the mixing.

I think you're losing track of some things tho. Reverb creates space, so adding reverb can actually make your voice sounds less stronger than you wanted. The idea behind Reverb is to simulate a room so you can fit your vocals properly in the mix, almost like you recorded them in a room with the desired effects. So Reverb is more something to fit the vocals in the song than necessarily something made to help with "strongness" and "loudness" of your vocals. Delay can make them big, sure, but it's like IO said, in context it will probably sound different (for both mentioned plugins).

Eq can help you in that, tho. By boosting the high end you can make your vocals sounds more present in the song (But becareful, it's too easy to mistaken with it), but again, try it in context because in solo it will probably do more harm than good. Believe me, I always tried to bring out CUL's vocal tone, but everytime I did a equalization it would either sound extremely the same I've done before or extremely different of what I wanted. When I understood it was because I was trying to equalize CUL's voice before even mixing her vocals I understood where I was mistaken.

IO already said, but I will say it too: use compressor, they're pretty good to make the vocal sound equally louder. Just compress it in a separate track (As IO said) otherwise your scream will lose too much loudness.

I will just advocate for you to use some type of saturator too because it will help you to make the track louder by adding even and/or odd harmonics to your sound, making lower frequencies higher what probably helps to create this loud scream sensation. Also important to note that to simulate a belting effect it can be important to have harmonics added to your higher vocal frequencies since this is how, "frequencially" speaking (XD), more or less belting works (And belting is nothing more than a singing scream well done).

Besides that, here are some tests I did:

(Saa scream)
Starts at 00:11

(Stop scream)
Starts at 00:08

I like the results pretty much! I think to replicate the "Saa" scream in the song you wanted you need to raise the pitch at the beginning of the note, since she sings the first "a's" more emphatically than the rest of the song, almost like if she was singing in an descedent bow. Something more or less like SAAAAaaaa. This is a pretty common thing in singing where you start or ends a note with a high and quick increase in the pitch. Unfortunately I forgot the name we give to that technique, but that's it.

Besides everything everyone already told you I think the only thing I can add for you now is to vary more that pitch at the beginning and/or end of your word, because this will empathise yor scream in the part you varied.

 

[mobius017]

Wow, thank you both so much for all the detailed advice! I'm looking forward to trying these things out!

So, things I'm learning:

• Don't mix in isolation
• Growl/Brightness/Clearness can be pushed harder for very brief periods, as long as the context of the song covers the distortions it introduces
• Try making two copies of the vocal track, one with Growl (and probably the changes to BRI/CLE) and one without. Compress at least the Growl track by itself. Then turn up the gain on the Growl track (while mixing in context) until it sounds right--meaning not too much or too little distortion.
• I was aware that Reverb/other effects can decrease the overall volume. Seeing it mentioned here made me think that that could be overcome with parallel processing, though. That would allow for keeping the gain while introducing the echoey sound, which might help reinforce the preception/illusion of a stronger yell.
• Boosting high frequencies with EQ can help the sound stand out. (Though Miku usually has plenty of high frequencies; I usually find myself trying to boost her lower frequencies and midrange, as well as adding saturation, to make her tone warmer.) Particularly for something like a scream, though, boosting the high end too could still be a good idea.
• Boosting both the low and high frequencies could be good for a scream.
• Try raising the pitch higher at the beginning of the note in the pitchbend.
• Try being the pitch more at the beginning/end of the scream. (Meaning vibrato/more drawn-in pitchbends?)

Do you have any recommended plugins for saturation? I have a couple (one with presets for lofi songs, and the saturation module in Nectar); sometimes I think they help, and sometimes I don't. Then again, I'm probably misusing them because I have the voice soloed while working with them, and I turn them up only until just below where I can hear the voice crackling. So I'm possibly not pushing them hard enough.

@AddictiveCUL (Add): Those samples came out great!

 

[IO+]

I think you got it all.

The idea behind this separate tracks is you have full control over how much growl you need, think of it as a Wet/Dry control. As for a growl track, if you plan to use it as parallel, please feel free to go nuts with it and you might try to increase an accent param to like 70-100%.
About EQing thing for growling parts, i usually listen for what it need but most of the time it end up with me using high shelf filter and pull 3-4kHz and above down quite a bit and using low shelf filter to pull anything below 500Hz up a bit this make the eq curve look like it tilt downward. After eq, i use clipper to shape the peak and in the mean time, it make waveform look like a fat sausage and it's sound quite fuzzy (depend on how hard you clip it) especially high frequency, this might explain why i high shelf it out around frequency that i mention. After that, i use compressor on top of it. It's depend on what it sound like after clipper, sometime i don't need to compress it further but if i use compressor i will pick something that have some color in it... It's like you put a salt, black pepper on top of the steak.

i'm not good at making vocaloids growling but this is how i do it, you might give it a try.

 

[mobius017]

Thank you! I did some work on it yesterday, putting parallel compression onto both the regular vocals and Growl part. It definitely helped with the gain I could put on the shouts, and let the Growl come in on the more stressed parts without wrecking the sound of the rest. I'm going to keep playing with the other items we've discussed here and see what I can get. I've shortened the notes and adjusted the pitchbends, too (might adjust the pitchbends some more).

Is a clipper an effect?

 

[IO+]

Is a clipper an effect?

Yes, they have very similar role to the limiter but they work differently, clipper is a waveshaper at the core just like any saturation plugin, while limiter are dynamic processor, mostly aim for transparency but clipper generate a lot of distortion and make thing sound aggressive while limit the dynamic, i find it's suit well for this kind of material.