NEUTRINO Tutorial - Revolutionary Vocal Service Using AI

Author: sleepfreaks

Software that uses AI to Create Vocals from Score Data


NEUTRINO is an application that has been garnering attention for it’s ability to turn notated score data (melody) into vocals (Kiritan or Yoko) using AI.


It doesn’t require detailed vocal edits, and creates surprisingly high quality vocals providing a revolutionary service.

It is currently only available on Windows, without an interface.
We had a number of inquiries as to how to make this work for a track.
Here we will look at the necessary steps from notating music to exporting to vocals to begin using Neutrino.

NEUTRINO Tutorial – Video Access

Preparing/Making a Melody File

Neutrino uses the MusicXML format to create vocals from score data.


MusicXML data contains note length, pitch, lyrics and other information relevant for a performance.


As mentioned in the Neutrino manual, we will be using the free MuseScore to create music xml data, but Dorico, Finale, Notion, and other notation software as well as some DAW can export to this music xml format.

  • MuseScore
  • Cubase

Download MuseScore :

We will open up Musescore to sequence in a simple melody.

After opening MuseScore, designate the key of the song.

Click on key to the left to select the key signature.
Next we will add in some notes.

Select note lengths from here, and enter them into the staff.
Next we will add in lyrics for the notes.

By going to the menu, add, text, lyric – while a note is selected

a text box will appear allowing you to enter lyrics, and we can enter a letter in hiragana.
Neutrino can only vocalize one letter (hiragana) per note.
Be sure to only enter 1 letter per note.

Next we will select the track tempo.

With the first note/rest selected, go to menu, add, text, tempo and a tempo display will be added.

We will stick with 120bpm.
Next we will export to music xml.

From the menu, go to file, export, and select uncompressed MusicXML as the file format.

We will save the file in the musicxml folder located in the score folder in Neutrino.

The file name will be used later, so we recommend choosing one thats easy to remember.


Designate the song tempo.


Enter the melody into the Piano Roll.

Cubase スコアエディタ

Select the MIDI event and select “open in score editor” from the menu.


Select the song key from “key” and click on a bar.


Click “lyrics” from “other”, and click on the note to enter in a lyric letter in hiragana.
Neutrino can only vocalize one letter (hiragana) per note.
Be sure to only enter 1 letter per note.


After entering notes, we will select the MIDI event.
With score editor open go to the file → MusicXML.


Enter a name and designate where to save the file.

  • MuseScore
  • Cubase

Entering a Melody File into Neutrino

Next we will look at the settings in Neutrino.

Download Neutrino from their website.
Right click on the “Run” file, and select “edit” to open the file in memo.


We will be changing this text data to change settings but first we need to get our MusicXML file that we exported.

Enter the MusicXML file name after “set BASENAME=”.
Be careful not to type in the “.musicxml” file extension in the name.


Neutrino contains the “Kiritan” and “Yoko” vocal libraries.
Change this here via “: NEUTRINO”.

Though the default singer is Kiritan, by changing after “set ModelDir=KIRITAN=” to “YOKO”, you can switch vocals.

“set FormantShift=1.0” allows you to change the vocal tone, and higher numbers give you a child like voice, while lower gives you a more adult like voice.
The manual recommends using between “0.85-1.15”.

Once this setup is complete, select save from the file dropdown list above, and close memo.

Creating Vocal Data Using Neutrino

Now we will finally look at creating vocal data.
First we will look at opening the Run batch file from the command prompt.

With the Neutrino folder open, go to the explorer address bar, and type in “cmd” and hit return.

This opens the command prompt; type in “Run.bat” and hit return to run it.
This completes our setup.
The vocal audio data will automatically be created while we wait.

Close the command prompt when you see the text “end”.
Lets take a look at “Output” in the Neutrino folder.

We can see that a vocal file has been created.
Neutrino happens to be free at this time, so we highly recommend trying it out and using AI to create vocals in your own music.