• We're currently having issues with our e-mail system. Anything requiring e-mail validation (2FA, forgotten passwords, etc.) requires to be changed manually at the moment. Please reach out via the Contact Us form if you require any assistance.

On voice (technical aspect)

user7465657

New Fan
Jul 25, 2020
2
Good day.
Could someone explain me please how vtubers change theirs voices? Does it need additional devices? I tried some software like voidol, voicemode, etc., but couldn't achieve my voice to be even approximately like theirs. Moreover some background noise always comes with my speech while I use such software.
 

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
Right now, the real-time conversion software Voidol doesn't really have satisfactory results and is more like a novelty (the thread peaches linked to is mine). I have only looked into Voicemod a little bit, so I can't speak about that software.

I watch VTubers and can understand Japanese, so I have actually researched a bit about VTubers changing their voices to become the opposite gender, which I assume you want to do. (Personally, I'm a girl who wants a guy voice, but I like watching male VTubers with female avatars.) When a guy uses a girl avatar, they call it babiniku (バ美肉). A male voice that has been edited to sound feminine is called babikoe (バ美声).

There are various pitch changing software, which are generally called ボイスチェンジャー (voice changers) that you edit your voice with after recording. (As a side note, Voidol is not actually a voice changer, it's a real-time voice converter that uses deep learning (AIリアルタイム音声変換ソフトウェア), that's why it sounds weird/has computery error sounds. But it did improve a bit with the most recent update in June if you haven't upgraded yet.) Another note, the way to achieve a feminine voice is a combination of pitch being higher and changing the formants (formants are basically vowels, male and female vocal chords are different, so when you convert a voice, the formants have to be changed properly to adjust the difference between male/female vocal chords or it makes it really obvious that it's fake.)

A lot of babiniku VTubers assist the pitch changing software by actively speaking in a higher, more feminine voice in order to achieve more convincing results. I really think if you haven't done so yet to look into YouTube tutorials by transwomen that teach vocal feminization/female voice training to get tips on how to practice making a feminine voice while not accidentally hurting yourself.

Anyway, I am going to list some voice changers the Japanese community uses. There might be English software, but I don't know about them. I have not tried any of these software other than Voidol, unfortunately. But if you have questions, feel free to ask and I can try to help/translate.
Clownfish voice changer (there are YouTube tutorials in English for this, it seems similar to Voicemod)


恋声 (koigoe) (have to download it by clicking the 詳細説明・ダウンロード button, scroll down to 「恋声」Ver2.87 のダウンロード)
I timestamped this tutorial's results and they seem pretty crazy!


ユーザーローカル (userlocal) The results sound kind of scary, there's audio from a tutorial around the 2:30 mark


I'm pretty sure there are other options, for koigoe is the most popular. They also use a DAW to further edit things, it looks like (EQ), but I'm not very good at using a DAW.

Oh, I forgot to mention, have you heard of NVIDIA RTX Voice? It can help with recording audio and removing any background noises. I have a GTX graphics card, so I was able to use this video to get around not having an RTX. But using RTX voice with Voidol REALLY did not work well. I think it would work well for voice conversion methods that take an already recorded audio clip, rather than real time, though.
 

user7465657

New Fan
Jul 25, 2020
2
Right now, the real-time conversion software Voidol doesn't really have satisfactory results and is more like a novelty (the thread peaches linked to is mine). I have only looked into Voicemod a little bit, so I can't speak about that software.

I watch VTubers and can understand Japanese, so I have actually researched a bit about VTubers changing their voices to become the opposite gender, which I assume you want to do. (Personally, I'm a girl who wants a guy voice, but I like watching male VTubers with female avatars.) When a guy uses a girl avatar, they call it babiniku (バ美肉). A male voice that has been edited to sound feminine is called babikoe (バ美声).

There are various pitch changing software, which are generally called ボイスチェンジャー (voice changers) that you edit your voice with after recording. (As a side note, Voidol is not actually a voice changer, it's a real-time voice converter that uses deep learning (AIリアルタイム音声変換ソフトウェア), that's why it sounds weird/has computery error sounds. But it did improve a bit with the most recent update in June if you haven't upgraded yet.) Another note, the way to achieve a feminine voice is a combination of pitch being higher and changing the formants (formants are basically vowels, male and female vocal chords are different, so when you convert a voice, the formants have to be changed properly to adjust the difference between male/female vocal chords or it makes it really obvious that it's fake.)

A lot of babiniku VTubers assist the pitch changing software by actively speaking in a higher, more feminine voice in order to achieve more convincing results. I really think if you haven't done so yet to look into YouTube tutorials by transwomen that teach vocal feminization/female voice training to get tips on how to practice making a feminine voice while not accidentally hurting yourself.

Anyway, I am going to list some voice changers the Japanese community uses. There might be English software, but I don't know about them. I have not tried any of these software other than Voidol, unfortunately. But if you have questions, feel free to ask and I can try to help/translate.
Clownfish voice changer (there are YouTube tutorials in English for this, it seems similar to Voicemod)


恋声 (koigoe) (have to download it by clicking the 詳細説明・ダウンロード button, scroll down to 「恋声」Ver2.87 のダウンロード)
I timestamped this tutorial's results and they seem pretty crazy!


ユーザーローカル (userlocal) The results sound kind of scary, there's audio from a tutorial around the 2:30 mark


I'm pretty sure there are other options, for koigoe is the most popular. They also use a DAW to further edit things, it looks like (EQ), but I'm not very good at using a DAW.

Oh, I forgot to mention, have you heard of NVIDIA RTX Voice? It can help with recording audio and removing any background noises. I have a GTX graphics card, so I was able to use this video to get around not having an RTX. But using RTX voice with Voidol REALLY did not work well. I think it would work well for voice conversion methods that take an already recorded audio clip, rather than real time, though.
Thank you very much for such comprehensive answer. Koigoe sounds great for sure, but its website wasn't that informative. Can I use it over OBS for livestream? Also, I've found this interesting video: My japanese is poor and I could hardly understand a quarter but as I got it the best way is combination of hardware and software for changing a voice (and VST plugins could be used just right in OBS I suppose). What do you think about it?
 

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
Okay, I ended up watching a short tutorial and I think I figured it out (it worked in both my OBS test and Discord tests). I only tested it long enough to be able to record my results and just dragged the pitch/formant slider to random amounts so that I could tell it was working, I didn't actually try to set it to a convincing/nice sounding voice. Note: I am using a headset with earphones and a mic attached, I'm not sure what your settings will have to be if you speak into a "proper" mic that doesn't have headphones, you will probably have to try out slightly different settings than me. I decided to use the headset, because I couldn't hear the output (my voice being converted) while using a "proper" mic and my OBS is already set up to record with the headset.

Anyway, I will write out the steps/show screenshots:

This is the tutorial I followed on how to use Koigoe in real-time on Discord and YouTube live streams.


SET UP KOIGOE
Click the microphone button under Input (入力) to convert your voice, press the square to stop recording.

Click the 設定 button for Settings in bottom right corner. (We'll come back to this step later.)
step1.jpg

You need to use a virtual audio device called VB-CABLE (Virtual Audio Device):
This software comes in English. Click the orange download button for Windows or Mac.
step2.jpg

Unzip VB-CABLE (if you don't have something to unzip files with, I use PeaZip, which is free). Locate VBCABLE_Setup_x64 (if you have a 64 bit computer, use VBCABLE_Setup if you have a 32 bit computer)
step3.jpg

Right click > Run as administrator
step4.jpg

Click "Install Driver" when the installer for VB-CABLE pops up.
step5.jpg

You have to restart your computer now in order for koigoe to detect VB-CABLE.

After restarting, go to koigoe's Settings (設定)...
Change Playback Device (再生デバイス) to "CABLE Input (2- VB-Audio Virtual Cable)".
step6.jpg

Change your Recording Device (録音デバイス) to whatever your microphone is (mine is "Microphone (3- Logitech USB Headset)").
Click OK.
step7.jpg

SETTINGS FOR OBS
Have the mic button pressed on koigoe (it makes it turn red) so it converts your voice.
step8.jpg

In Settings for OBS (bottom right corner), select "CABLE Input (2- VB-Audio Virtual Cable).
step10.jpg

Just press Start Recording and everything should be working properly. (I didn't test with streaming.)


DISCORD SETTINGS
Click the gear icon to get to Settings.
step11.jpg

Go to the Voice & Video tab. Select your Input Device as the virtual cable ("CABLE Output (2- VB-Audio Virtual Cable"). Select your Output Device as your microphone (mine is "Speakers (3- Logitech USB Headset").

You can press Let's Check to test out hearing your converted voice.
step12.jpg

@user7465657 Hopefully these steps help you so you can test out koigoe. Please let me know if you need me to research anything else. (I know I didn't really look into the settings for the voice, maybe I will do that after you get the set up for Discord/streaming working? I was kind of busy today, but finally managed to squeeze in time to write this.)
 

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
Here is a translation of the main interface/settings. I will update the list if there are better translations when I look over the manual/test some things out.

恋声 / koigoe / Love voice

方法の選択 / Chose method
TD-PSOLA (単音向け / For monotone)
Phase Vocoder (複数音向け / For multiple sounds)

入力 / Input
一時停止、停止 / Pause, Stop

声の高さと性質の調整 / Adjust voice pitch and quality

ボーカルの音域 / Vocal range

出力音声をファイルに保存する / Save voice output as file

歌声 / Singing voice
話声 / Speaking voice
3度上を重ねる / Stack 3rds
平坦化 / Flatten
正弦波 / Sine wave
方形波 / Square wave
鋸波 / Sawtooth wave

設定 / Settings
=====================================
恋声の設定 / Koigoe Settings

共通の設定 / General settings

録音デバイス / Recording device
再生デバイス / Playback device

音声遅延 / Voice latency

A4の周波数 / A4 frequency

出力WAV,Dumpの保存先 / WAV output,Dump save destination

EcoDecoTooLの場所 / EcoDecoTooL location
設定なし / Not set

MIDI 出力の Instrument No. / MIDI Output Instrument Number

TD-PSOLAの設定 / TD-PSOLA Settings
バッチ処理のオーバーサンプリング / Batch processing over sampling

音量のグラフ表示を行う / Display volume graph

キャンセル / Cancel
 
  • Like
Reactions: rexzone

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
Tutorial for Koigoe for real-time voice conversion in Discord

YOU WILL NEED:
1) Koigoe: 恋声
Scroll down to 「恋声」Ver2.87 のダウンロード / Download "Koigoe" Ver2.87

2) VB-CABLE: VB-Audio Virtual Apps
The website is in English, click the orange download button. The driver pack contains two exe files, VBCABLE_Setup.exe or VBCABLE_Setup_x64.exe. The tutorial recommends right-clicking VBCABLE_Setup_x64.exe & running it as administrator > Click the Install Driver button (when you click this, it makes your audio temporarily stop working, if I remember right, you have to restart the computer)


Tutorial source:


0:00~1:00 of tutorial explains that this is the normal way to use Koigoe, but it won't work with Discord.
Click 設定/Settings on the bottom right corner
録音デバイス/Recording device = the mic you are talking with
再生デバイス/Playback device = your speaker
The chain becomes mic --> Koigoe --> speakers
normal.png

1:00~3:00 = installing VB-CABLE
But if you want to use it with Discord, you have to have a 2nd mic, but that would be expensive & annoying.
So use a free virtual cable (they recommend using "VB-CABLE Virtual Audio Device").

3:00 = How to select the correct settings in Koigoe to prepare for Discord use
録音デバイス/Recording device = Don't set this yet, select the playback device 1st (selecting playback device resets what you selected for recording device so you have to do it in reverse order), I will explain in order, so please just keep reading.
If you right click the volume slider on the Windows taskbar > Sound > you will see "CABLE Input" in the Playback tab toward the bottom, this is what installed with VB-CABLE.
playback.png

再生デバイス/Playback device = If "CABLE Input (VB-Audio Virtual C" is not on the Koigoe playback device list, please try closing/restarting Koigoe.
録音デバイス/Recording device = From the drop-down, select the mic you are speaking through (his is called "1-2 (QUAD-CAPTURE)", yours will most likely be a different device!)
virtual-cable.png

Before clicking OK, change the radio button like this:
radio.png

He got an error message for both settings with the radio button, so it seems the error can be ignored:
radio-error.png

To check if your setup is working, click the mic icon --> this will make blue sound waves appear in the Pitch Analyzer --> In Sound, the green bars will fill up to mean sound is being registered.
setup-ok.png

4:00 = Discord connection
His default Discord settings are for his mic (called "1-2 (QUAD-CAPTURE)", yours is most likely different).
default.png
He clicks "Let's Check" and his voice is echoey but sounds like himself/unedited by Koigoe.

TO BE CONTINUED IN NEXT POST
 

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
To get your voice converted by Koigoe, change the Input Device to the virtual mic "CABLE Output (VB-Audio Virtual Cable)" & then press the mic icon in Koigoe:
discord.png

Make sure that your setup is working in Koigoe (blue sound waves appear in the Pitch Analyzer --> In Sound, the green bars will fill up to mean sound is being registered).

4:53 In Discord, click "Let's Check". He realizes it doesn't sound right because the settings are default.
default-voice.png

Press the stop square button in Koigoe to halt voice conversion. He clicks the M --> W preset (man to woman preset), but it is still not good enough.
Note: If you have a feminine voice and want it to become masculine, please try the other settings, such as W --> M.
m-w.png

5:08 Press Effect to fine-tune the settings.
The only thing he edited is adding 2 check marks for EQ and Reverb:
effect.png
Click X to close out of the settings to move to the next step.

5:16 Click the mic icon in Koigoe and make sure to press Let's Check in Discord. His voice now sounds higher-pitched and is successfully converted.
done.png


Questions answered by the tutorial-maker:
1) Can I use Koigoe in Fortnite's voice chat? --> If you change the recording device within Fornite's settings.
2) It's not working!! --> No sound from speakers is normal (because sound is being output through the virtual cable), but if the meter in Koigoe or Discord is not moving, there is a problem. Check your settings to make sure the microphone is functioning properly and also that you have selected the input device properly in Koigoe/Discord.
 
Last edited:

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
Settings shared by a VTuber called Yua to assist with masculine to feminine voice:


Pitch = highness of voice. High number makes the voice high pitched, lower number makes the voice lower pitched. The tutorial sets theirs to 126%.
Format = changes voice quality. They described a high number as making you sound like a woman getting an interview on TV, a low number makes you sound like a criminal giving a testimony. They set theirs to 126%.
01.png

In addition to changing these settings, they recommend mimicking your ideal voice to help the conversion go smoother. (Because it's not a perfect tool.)
This will require practice speaking in a higher voice.

To find your ideal settings, start with the default 100% and slowly change the sliders.

Click the Effects button to get up the EQ and Reverb settings.
02.png

EQ lets you adjust volume of pitch (louder or softer). The VTuber prefers to raise C3 and C4 slightly above the default.
If your environment is noisy, adjust C0 and C10.

Reverb adds an echo like during karaoke. If it's too high, it makes the voice seem unnatural. But if you raise it a little bit, it gives a nice atmosphere, so they recommend this.

Lastly, Yua adds a filter in OBS. (OBS is free and used for streaming.)
They use the following 3 filters:
ノイズ抑制 (noise suppression) set to -5. This reduced noise. The number you pick depends on your environment. If you change it too much, the voice quality will suffer. Probably don't set it more than -20.
noise1.png
How to access filters in OBS = click the gear icon --> Filters
noise4.png

Click the + icon to add your chain of filters:
noise5.png
English tutorial about noise suppression in OBS:


ノイズゲート (noise gate) set to -70, -65, 10, 200, 300. It's an effect that cuts small sounds that are louder than the volume you set. Test out the numbers to see what you need.
noise2.png
English tutorial about noise gate in OBS:


コンプレッサー (compressor) set to 11, -16, 302, 99, -1.30, none. It's an effect that makes the volume of the voice even/fixed.
noise3.png
English tutorial about the compressor in OBS:
 

Users Who Are Viewing This Thread (Users: 0, Guests: 1)