I don't believe this is real because growl is NOT just a parameter, it requires new recordings. They literally had to find a new VP to make them so they could port miki and Yuki.
In the v4 dev kit there's stock growl samples. Internet co v4 updates all use these stock samples. Apparently it is really hard to record so most banks use the stock samples.This has been confirmed by Dex and Daina's voice providers.
This is true, but there's no way to add growl samples to V2 banks.
I guess it depends on what your barometer for "better" is. Just in terms of audio output? Yeah, NNSVS still has a long way to go before it's of comparable quality with any of the big hitters. But for some people (like me tbh...), just the fact that it's free and open source does make it better in certain ways.
I've seen people compare the audio output and their reasoning is "it's free and open source". Those are two different comparisons...
I think Miku has her own growl (if you go into her DDB there's a sample that sounds like her growling) and I wouldn't be surprised if Fukase JP Normal did too, considering how his growl sounds
Quadimension's ai plus banks aren't trained on the tuned renders of the synth v voice banks, rather they are trained on the recorded samples that make up the voice bank. Thinking about it as encoding the voice bank in ai rather than standard synthesis. I hope that makes sense. To my knowledge there hasn't been a released ai singer trained on tuned vsqs. With how crypton likes to reuse Miku v2s original recordings for all her voice banks I wonder how an ai trained on it would sound. I feel like it might be the way to get the purest form of Miku because its not being distorted by a sampler. I honestly don't think nt sounds like any of the original characters they all sound distorted in a unappealing way that makes it hard to use them over their vocaloid version.
Ohh, I see. I thought it specifically had to be singing samples so I assumed otherwise
yea, they were similar and the power appends were thought of as higher quality so once the Kagamines received them in V2, that's what producers predominantly used.
people say that nnsvs is better then sv, cai and v6? lol, i have only seen people shit on nnsvs
I'm fairly certain they're trained on a mix of the original samples, untuned renders from SV Basic, and other filler data.
I see. I guess in Miku Light and Vivid's case, Original and Solid can fill their roles pretty well anyway?
I don't think so, personally... I miss Miku Vivid and Light so much;;;; Miku Solid is excellent for a more grounded tone, but nothing can replace the brightness and liveliness of Vivid and Light.
And I'd say Vivid is such a fitting append (like Solid is and so does the rest) I feel like if they were ever do an AI version of Miku using the same trick as Quadimension did with the MEDIUM5, they should really rely on the recordings used for Vivid
What if they did all her appends though?
Now that I think of it I do remember that anime characters tend to be hit by trucks fairly frequently
Eh, honestly, that’s an incredibly common opinion regarding CeVIO. I don’t think anyone’s denying it’s noisy, it’s popularity simply comes from people finding that they like certain things about it more than they dislike the noisiness. (Unpopular opinion of my own: I thing CeVIO’s noise has gotten a lot less bothersome the further along it’s come. Compare Yukari Rei to Sasara, for example; Sasara and everyone that’s come after her gave noise, but it’s not grating and invasive like Rei’s is.)
