Back to top of page
The Ultimate Vocaloid Growl Tutorial
Black Metal Miku.

How to make a VOCALOID growl / Death Metal scream / Black Metal shriek. This is my personal method to get a full sound, without focusing on how realistic it is. I'm 90% sure that this is how Utsu-P got his main vocals 'till Algorithm, after that, they got clearer and less processed, but the first few steps still apply. I'll explain how to replicate them with VOCALOID 4 and 5, but the methods for V4 also work for V3.

Also I've included both project files, .vsqx and .vpr, as per the final render of my song "Hypocrite", so you can compare the clean and the original version and see which method you like the most. The download will be at the end of the post.

Introduction and alternatives

I'll start by saying that there are vocal synthesizers that have this sort of feature built-in, and that it is way easier to use them instead. Namely UTAU Hakaine Maiko and Synthesizer V.

Hakaine Maiko.
Hakaine Maiko. Source:
SynthV Glottal Effect Editor.
SynthV Glottal Effect Editor.

If you still want to turn cute Hatsune Miku into a grueling guttural diva, power through, cause it is a little complicated and rewards don't come till the end.

The most important step to get a proper end result, is to work with the best source you can, just like recording a real instrument. What we want to emulate is called fry scream, you can learn about it here, though it is not neccesary, it helps to understand what we are dealing with:

What I will teach you, is how to get that static or white noise sound that Mary showcases, then you can distort it into a full scream however you like, for example with a virtual guitar amplifier, a dirty compressor, etc.


This works on VOCALOID 3 and I believe it should work on Piapro Studio, but I haven't tried it yet.

The trick is to use the voiceless phonemes that the engine can generate since V3, so any voicebank can use this. To do so, you only need to append _0 to every phoneme. Just right click on any note and select Note Property, then in the PHONETIC field you can input the change.

For example: [ D V ], would become [ D_0 V_0 ].

Voiceless Phoneme Edit.
This needs to be done for every single phoneme.

This is extremely time consuming, luckily, there's a job plugin that does this for you: WhisperAll.lua. Sadly the official VOCALOID Store that hosted most job plugins is dead. Luckily I have most of them and here's a download link to them:

To use the plugin, simply go into the VOCALOID part you want to modify and run it.

Whisper All Job Plugin.
You can set and unset the _0 whisper phoneme automatically.

Now we have a vocal that sounds like this:

Note however, that consonants are not treated the same way as vowels when devoiced, they will be much louder than them, especially plosives. So you have to edit them now, or they will become super annoying after we distort them. You can level the vocals either in the VOCALOID EDITOR or in your DAW.

Consonant Edit.
I find it easier to control the DYN parameter instead of messing with volume automation in the DAW.

Our next step will be to add some of that gargling sound that is so characteristic of vocal fry, though I will show you how in another section, you can skip to it by clicking on the table of contents on the side.

Now, how to create this whisper with the VOCALOID 5 EDITOR.


This is actually a lot easier to achieve with the new editor, as they've added a new STYLE tab with everything you need, the most important part is the VOICE COLOR section, for now, you can just copy the settings, it is easy to figure out what each slider does by playing with it.

VOICE COLOR settings.
ROBOT VOICE and BREATH also contribute to the sound in a big way.

Afterwards, we only need to process it, unlike with the V4 EDITOR, we can do some serious weight lifting in here:

AUDIO EFFECT settings.
Again, copy them, then adjust to fit you particular vocal.

Some explanation may be needed here:

DE-ESSER: Think of it as a compressor for "S's" "T's" and other high frequency plosives, it will turn them down automatically so you don't have to. It is important to adjust the threshold 'till you get a decent amount of GAIN REDUCTION.

GAIN: After our VOICE COLOR adjustment and DE-ESSER we lost a lot of volume, as the VOCALOID AUDIO EFFECTS don't have an input knob, we must dial in the gain in this step.

TREMOLO: Used to get a little of the "gargling" sound of vocal fry, we'll use another plugin for this once we get to editing in our DAW, so only a mild amount of tremolo is needed now.

EQUALIZER: Pretty much all we have at this point are highs, to avoid a really harsh vocal, we can shape the tone a little now.

DELAY: Just for previewing while working on the song, I tend to remove it once I export, as I prefer my VST delays.

CHORUS: Same as for the DELAY.

Now we should have something that sounds like this:

Notice how much softer and controlled it is when compared to the V4 version, not that you can't edit the .wav to get the same results, but all you have to do to get this far is slap the style preset and you're good to go.

Distorting a quiet whisper

You still with me? This is the fun part, it is also where every producer does things differently. I'll show you 2 vocal chains, one for each method above. We'll be using our DAWs from now on, I personally like to use REAPER, but you can use whatever you like.

There are 2 VST plugins that are essential for you to get a similar final file, you can replace all the others with whatever tools you happen to have.


The developer of this plugin has long since closed his website, I have a backup of all his free downloads. You can get only charsiesis, but I suggest you play with all of them as they are very unique.


This one was developed especially to get this kind of growling effect.

V4 Vocal Chain

The first thing I will show you is the routing setup, for this I've divided the .wav straight out of VOCALOID 4 into 3 tracks where I can process each one individually. You could always just triplicate the orginal track, but you save on plugins and thus processing power when you do it this way.

Audio Routing.
Disable Master Send for the first track so that you only hear the tracks you are sending to and not include the original.

The cool thing about send/returns is that you can alter the source and thus every track you are sending to, only by adding plugins to 1 track. In this case, I've added a limiter plugin to squash all the consonant peaks I might have missed when editing.

Limiter Plugin.
A limiter is essentially a faster, more agressive compressor, it is used to "limit" the amplitude of the loudest peaks of audio. You want a file that looks (I should say sounds) like the upper track, not as spiky as the bottom one.

Then this is how I processed each track:

Growl Miku Hi:

Growl Miku High.

First in the chain is VocaScreamer, this is a wild plugin so try to use mild settings.

Next one is Charsiesis, this is a complex effect, you can think of it as if you split your signal into multiple delay pedals, the voices knob controls how many you split into. Then imagine that there is a dude playing with the knobs of every single one of those delay pedals.

The min delay and delay range control how much they can turn the delay time knob.

The min rate and rate range control how often they turn the delay time knob.

The rate update is a technical implementation, the plugin must update its parameters internally at this rate.

The other parameters don't really interest us as they have to do with filtering and modulation, though they are cool to play with it if you put the plugin on a synth or other instruments. If you want to learn about them read the included manual.

For the purposes of this tutorial, these are the settings that work best:

voices: At 16 or 32

min delay 2ms

delay range 20ms

min rate 240ms/s

rate range 285ms/s

rate update 50 units

feedback 21%

mix 35%

The EQ is self explanatory, it pretty much replicates what I did on the VOCALOID 5 AUDIO EFFECT section.

This is how Growl Miku High sounds soloed:

Growl Miku Mid:

Growl Miku Mid.

This one is more straightforward, I've simply added a pitch shifter before going into VocaScreamer.

It sounds like this:

Growl Miku Lo:

Growl Miku Lo.

This is a sub-bass track. Pitch shifter into VocaScreamer into Charsiesis.

I run slightly different settings in here:

voices: 32

min delay 2ms

delay range 20ms

min rate 0ms/s

rate range 10ms/s

rate update 50 units

lowpass 53%

feedback 0%

mix 100%

pre 7.5db

Then in the parent folder I've shaped the tone some more, you don't have to use these plugins, if you have them cool, if you don't, use any EQ.

Oneknob phatter simply boosts the mids and distorts them a bit.

Growl Miku Summing.

When you sum it all you get this:

It may not seem like that much of a difference compared to the first track alone, but it does make an important one when you post-process this, which I will get into in the next part.

V5 Vocal Chain

Alright we are almost there! If you ignore all the plugins before VocaScreamer, you can use this exact vocal chain to enhance what you wound up with using the V4 method.

First we clean up the low end as we have our own source of controlled rumble like I've showed you before, a parallel track with a pitch shifter.

LinEQ Lowband Stereo.

Then some general purpose EQ, once more it doesn't have to be this EQ, I just like the API for it's simplicity.

Remember that it is always better to clean a track, that is, to reduce problem frequencies, before boosting.

API-550B Stereo.

That leads us into the next step, boosting, this time with a compressor, you generally want a fast attack and a strong ratio to tame all the peaks. Think of this one as a softer limiter like the one I showed you before.

API-2500 Stereo.
You can hit the API pretty hard! Around 12 db of gain reduciton is 0k!.

You already now the next pair ;)

The settings this time are:

voices: 16

min delay 2ms

delay range 20ms

min rate 240ms/s

rate range 285ms/s

rate update 50 units

feedback 21%

mix 35%

post 2.5db

Charsiesis and VocaScreamer.
Charsiesis and VocaScreamer are a match made in heaven.

Afterwards another compressor! I like to use multiple ones as you don't have to push them as much, besides we added a lot of noise with the last 2 plugins.

DensitymkIII is a free plugin, you should definitely go download it as it is absolutely amazing.

Density MK III.
Try to get around 3 db of gain reduction with this one, as it's only meant to do light leveling.

Next is a DeEsser, some vocaloids needs this at this point, others maybe not, don't just blindly follow along.

DeEsser Stereo.
6 db of attenuation should be fine.

Pultec! I'm only using the high shelf, be wary of killing all your top end.

PuigTec EQP1A.

Oof I'm kinda tired at this point. ReaDelay is free too, comes with REAPER. I'll link you to this preset, as it's easier than making 6 screenshots.


One last touch before we go.

FabFilter Pro-Q 2.

So all of that for what! Judge for yourself:

The delays sound kinda insane, but in a project with a lot of guitars, it is actually a very reasonable amount of wetness.

Now there's a couple things I haven't showed you here, but they are minor things, like a reverb where I send into and some stereo spread happening in parallel, but this post is waaaaaaaaaaaay too long already and I highly doubt you'll make it this far. If you have done so, please show me what you made with it. I hope it served you well, because damn I would have payed for something like this when I started playing with VOCALOID.

Bonus round: Black Metal shriek

You might be amused by the fact that I don't use the GWL(Growl) parameter, that is because GWL only affects pitched notes. When you use the voiceless phonemes, all the sounds you produce are essentially pitchless, so they are not affected by GWL automation. However, where you place the notes on the piano roll does change the sound, it tends to distort it in different ways, don't be afraid of moving the notes to the very extremes of the MUSICAL EDITOR, as suddenly the grayed-out areas become very interesting.

Having said that, in this part I will show you a method that does use the regular pitched samples, it exploits both the GWL parameter and the distortion caused by placing notes beyond "C5".

This will only work on V4 voicebanks as you need the newly added GWL samples and not everyone will behave the same, some go full robot mode when you push them past their spec. Miku V4 is lovely for this trick.

All you need to do is max out GWL and BRE on a really high pitch, you can use DYN and PIT as volume and tone controls. What you get is a wailing screech a la Dani Filth, really wild stuff!

Growl parameter.
Breathiness parameter.
(Breathiness parameter, think of this as a softer whisper hack.

Here's how it sounds:

When you go towards the other end, below "C1", you get something that resembles an actual vocal fry, it is very noisy though and I've never gotten anything useful out of it. I do know that HERNIA CROCO has done so. If you play enough with it, you can figure something out.

Here's a link to one of his covers that showcases this kind of extreme vocals:

Hypocrite VSQ download

As promised, here's the .vsq for my song, if you make a cover with it, I would love to listen to it <3.

This one doesn't have much tuning going on, if you can, always KISS (Keep It Simple, Stupid). Kinda ironic after all this typing, welp.

If it complains it can't find the instrumental .wav, you can download it for free on my Soundcloud page.

Thanks for reading! Have any ideas? Please leave your comments, suggestions, or anything that can add value to this post down below.

Also please follow me on social media, so that you don't miss any updates on my music, posts or videos.

Have fun.

Twitter //
Youtube //
Bandcamp //
Spotify //
Soundcloud //

Welcome to my comments section!

We do things a little different around here. To post a comment you have to mail it to:

And use this post title as the subject line and nothing else, otherwise it will go straight in the trash:

The Ultimate Vocaloid Growl Tutorial

Why make it so difficult?

Because this is a static site. I have no need for fancy server-side programming. This does mean that posting a comment can take a few minutes. If you want to chat with me in real time you can find me in twitter.


Why not use Disqus?

Because like most other comment providers it does not respect your privacy and actively sells your data.

I hope this is not too incovenient for you, please comment if you have something to say, I value it a lot <3

Before you mail, ask yourself what would Tohsaka think of your comment?


名前:(*゚Д゚)さん:2006/01/18(水) 19:46:45 ID:ANUP1kKV
         ヽ:::/     /::::::/_,,... -‐‐‐- 、 _
          >!      l:::::シ´ _,. -‐ ''" ̄` `ヽ、
        /;'     ...!::::::ヽ、´  ,     ヽ `ヽ、
       く::::::! .    .:::|____r' / /  /   ハィ⌒ヾ、
        >::! :    ::::l::::::l / ///、,ィ' /!   /、,,:;!i   ヽ
        ヽi :    ::::レ'ヽ/ l/_L/,_ト/ ! ,イ l  l:l:   ,`、‐-、
         l :    ::::l l 〉'   '`7-ヽ.`iX !|  / , ;i.  i:.. l::::::::>
         | :     ::l (    `ヾ;シ^'   __,ノ_ノ/::! , l::::レ'´
         l :.:    :::し,          ´ィ=z、_7///:::/!
         l: ::    ::::::!i         、 ゞ-',.゛イ〃:/::l
         ! :::.    ::::| ヽ、     _ ...._ ´  /  i :l ::! ̄´
        / l ::    .::::|  \   ` ー '  , イ  | .| :|
       / / :    ::::::|    \    ,. イ::::|   l .l ::!
      /  / .:    :::「丁!‐‐‐「! ̄ ̄`!::::::l:::i:::..:  l .l ::l
     / .:/ :::    :::| l l   l l  _,,,_l:::::::リ:::   / i:::l
    / ..:::/  ::  ____,;;シ'_//,___ l_l/ ............´  /  /::::l
   / .:::∠,-r--r-|----ァ ハ r-、`ヾ .:'   .............//ヽ::::i、
  / /   l  L二// l ! l\ \ ...:':;: -‐‐‐ '´   `、ヽ
/  i    丿  //  ,,.l l- l ヽ/ ::::/         ∧ i
    l/ /   く/ ヽ/ l l  |  ! ::::|         / l !
フェイト ステイナイト 遠坂凛