What's new

Synthesizer V - Vocaloid haters might want to check this

Pier-V

Is doing a barrel roll (actually an aileron roll)
I recently discovered that one single guy is basically writing the future of vocal virtual instruments, but despite the huge fanbase none of the "pros" seems to have noticed.
To be more specific, this company named Dreamtonics is working on a Vst, Synthesizer V. This Vst uses voicebanks, and the one that really captured my attention are Saki AI for japanese and Eleanor Forte Lite for english (the AI version is still in development).
It's basically like Vocaloid, but with the big difference that it's not a meme - yeah, maybe Porter Robinson made Avanna work but it's the exception to the rule (btw I proudly own Avanna... sorry Avanna but it's true).
Every voice in the excerpts below doesn't have a real life counterpart, and while Eleanor still needs work I have to admit Saki is impressive.
However, since neither japanese nor english are my native languages (as you can probably tell) I am genuinely curious to know what you think about all of this.
One last thing: listen the first two pieces as you like. The third one... please just don't use headphones.


 
I did not pass the test in the third video. Thank you for the warning.

I’m very impressed. Just deciphering the word is usually a challenge, but there were very musical. The second one was a bit lifeless, but almost right for a a certain kind of affectless style.

Thanks for sharing these.
 
Where did you get the only englisch voice "Eleanor Forte lite" ? I looked on the dreamtonic site and it says "to be announced" only.

I definitely did not need any japanese or pikachu voices ;)
 
Currently, the only English AI Voice released is Tsurumaki Maki, voiced by a japanese voice actress. ANRI and Eleanor Forte AI are due to be released by the end of the year.

Also, there was a successful indiegogo campaign for SOLARIA - a synthesizer V AI voicebank by Eclipsed Sounds, voiced by nashville based singer Emma Rowley, due to be released in Feburary 2021.

There's also a male voicebank in development.

As for limitations with the lite version :

lite Standard voicebanks contain only one pitch of recorded phonemes, whereas the full version has multiple pitches of recorded phonemes.

Lite AI voicebanks render in a "speed" mode within Synthesizer V, which creates more noise than the "quality" rendering mode.

Lite voicebanks are for noncommercial use only.
 
I find it fairly 2d and emotionless in both languages.

I'm sure it will be popular though but in only replacing 3rd tear pop artist. Certainly isn't going to be replacing singers like Adele or Ariana Grande or Demi Levato.

I'm old school I guess. I like my singers to have SOUL!!!!

 
I find it fairly 2d and emotionless in both languages.

I'm sure it will be popular though but in only replacing 3rd tear pop artist. Certainly isn't going to be replacing singers like Adele or Ariana Grande or Demi Levato.
Gosh no! It isn’t going to replace live recording of real singers at all. Other than the novelty market, I see this as being good for replacing or augmenting the use of sampled vocals and some vox-like effects. And making demos or mock ups. I’m sure that some will be hoping for more out of it.
 
Currently, the only English AI Voice released is Tsurumaki Maki, voiced by a japanese voice actress. ANRI and Eleanor Forte AI are due to be released by the end of the year.

Also, there was a successful indiegogo campaign for SOLARIA - a synthesizer V AI voicebank by Eclipsed Sounds, voiced by nashville based singer Emma Rowley, due to be released in Feburary 2021.

There's also a male voicebank in development.

As for limitations with the lite version :

lite Standard voicebanks contain only one pitch of recorded phonemes, whereas the full version has multiple pitches of recorded phonemes.

Lite AI voicebanks render in a "speed" mode within Synthesizer V, which creates more noise than the "quality" rendering mode.

Lite voicebanks are for noncommercial use only.
What does "one pitch" mean, only one note o_O ? That would be absolutley useless.
 
Gosh no! It isn’t going to replace live recording of real singers at all. Other than the novelty market, I see this as being good for replacing or augmenting the use of sampled vocals and some vox-like effects. And making demos or mock ups. I’m sure that some will be hoping for more out of it.
I guess my views were tainted when I watched a video of 1000's of little Japanese kids crying to a hologram and vocaloid. Have people lost so much touch with what it means to be human that this kind of "performance" is at all appealing? I mean for a generation that grew up on "fury" stuff maybe that's what it's come down to.

But I do see potential in Anri. I use to do that with vocal phrase libraries and melodyne. It was pretty fun. seems like Anri with the right recordings could actually be something.
 
Last edited:
What does "one pitch" mean, only one note o_O ? That would be absolutley useless.
standard voicebanks, which are not really being produced for the engine much anymore, are recorded by the voice providers singing a string of all possible phonemes for the given language, one pitch at a time, for how many pitches they choose to record, and is then pitched in engine accordingly, like a sampled instrument, so the further you go from the pitches it was recorded in the tone/legibility deviates more and more. Since the lite version only contains one pitch instead of the multi pitched full version, the tone shifts a lot more. Think a Piano Library with 127 velocity layers but only one note sampled so the tone falls apart outside of a small range.
 
the lite AI voicebanks are the same as the full voicebanks, just with faster, lower quality rendering in engine

Also, to make sure you all know, the first example in japanese was manually tuned/manipulated by hand, and is not what you would get right out of the program, though the AI voicebanks do have an auto pitch mode that creates pitchbends based off the way the voice provider sings
 
thanks for that hint! I didn't expect anymore, that the artificial vocalist technology will develop into a serious direction after all of the silly smurf voices of vocaloid. sounds promising!
 
Top Bottom