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.
Saki AI was really the only AI vb that got my interest just because I liked her standard SynthV voice to begin with. Although idk if it's just me but all the uses of her I've heard all sound like they've been tuned by the same person, which isn't to say its BAD tuning it's just something I've noticed after hearing covers with her.
I'm currently working on a Saki Ai project (and by currently I mean SynthV is open in another window as I type this). The AI pitch data still needs tweaking here and there, especially for downward leaps, so it's not fully automatic. Plus, the random seed feature is really handy. The strategy I've developed is to highlight an individual phrase and then keep selecting "Rerun with a New Random Seed" until the tuning is where I like it to be. Once I'm happy, I move to the next phrase and repeat the process.
Something I’d like to see with AI banks would be people who do post-render editing with them (by which I mean render the file with the AI and then edit/refine the tuning using something like Vocalshifter).
If that sounds interesting, iirc, I think @haru0l doss something like that. (That, or Haru uses AI tuning sometimes to tune UTAU/Vocaloid via Vocalshifter. I can’t remember, lol.)
Hey, I’d never thought about that! But still, like... there’s only so much you can do and so many results you can get with an AI bank. They’re not nearly as flexible, y’know? Working with them just isn’t as satisfying once the novelty wears off.
With SynthV you can freely edit and/or draw over top the AI-created pitch data, so in that respect the AI tuning can be used as a fancy-shmancy starting point. Of course, you can use Saki AI without the AI tuning function and create whatever freehand tuning you like, while still benefiting from the less harsh sound of Saki AI.
I think the point that's being made here is not that we don't acknowledge that these are features that can be used on Saki AI but instead when they are executed, due to the nature of the vb and how it reacts, it doesn't end up sounding nearly as personalized, like how it was noted by some of us that she always sounds like tuned by the same person, we could either have the ai tune it or we could tune it manually but the end product won't end up being drastically different either way, which is not satisfying. So it's more about the vb's behavior as a whole.
I think it is because we are getting a lot closer to singer in a box territory and less complicated instrument. I think it's also the type of voice banks we are getting and how these engines effect those sample kinda like how cevio has a unique sound to it.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.