NEUTRINO Neural Singing Synthesizer Available for Download
On February 22, 2020, SHACHI unveiled the NEUTRINO Neural Singing Synthesizer to the world, and has now made it available for download!
About NEUTRINO
NEUTRINO is a freeware voice synthesizer program that reads notes and lyrics from a MusicXML file (*.musicxml) created in a musical score program such as MuseScore. Using neural networks, it can estimate parameters like timing, pitch, voice type and more. It then uses a vocoder to synthesize the voice that sings the notes based on those estimated parameters.
A sample can be heard here:
What’s more, two different voice banks are already available. The first is 謡子 / Yoko, a voice bank based off the singing voice synthesis system Sinsy, and described as possessing a powerful yet cool and relaxed voice type. Below, you can hear song that SHACHI created for Sinsy’s recent 10th anniversary!
The other available voice bank will no doubt please AHS fans. It’s none other than AI Tohoku Kiritan! Below, you can listen to numerous sample tracks that SHACHI created using her voice database.
More samples can be heard on SHACHI’s SoundCloud page.
Interested in hearing more of her? You’ll be delighted to know that many users have already tried their hand at Yoko and Kiritan’s NEUTRINO banks, and have shared their results online! Below are just a few of the many examples out there!
Many more user songs can be found by searching on Niconico, or searching on Twitter for #NEUTINO, #AIきりたん, #AIシンガーきりたん, or (of course) Tohoku Zunko’s official Twitter!
Download
Impressed by what you hear? You can download and give NEUTRINO a try for yourself below!
Testing
After unzipping the file to your desired folder, open the resulting NEUTRINO folder, double-click on the Batch file called “Run.bat”, and let it run for a while.
When it finishes, check the folder called “output” and you should see a WAV file called “sample1_syn.wav”. This means that the program works and you are ready use it!
How to Use NEUTRINO
1. Create a song with lyrics in MuseScore. Save it as MusicXML file (*.musicxml) in “score” > “musicxml”
※If you see 4 sample MusicXML files, then you got the correct folder
2. Right-click the earlier Batch file called “Run” and open it in Notepad. Change where you see BASENAME into the name of the MusicXML file that you want to synthesize.
※”sample4″ is being used in this example
This is also the step where you can adjust the parameters (PitchShift, FormantShift, etc.) to your liking, or change Kiritan to Yoko. SHACHI gives a brief tutorial about how to do each here (Japanese only).
3. Launch Command Prompt in the current folder. You can learn how to do so here.
※The easiest way is to type “cmd” into the folder address bar, and hit Enter
4. Enter “Run.bat” in Command Prompt, hit Enter, wait a little while, and then the newly synthesized WAV file will appear in the “output” folder!
Recommended operating environment
Operating System: Windows 10 (64bit)
CPU: Intel Core i5/ AMD RYZEN 5
Memory: at least 8GB
Free disk space: at least 10GB
Keep up to date on this exciting new voice synthesis program by following SHACHI on Twitter and visiting the NEUTRINO official site!