I personally thought the majority of XLS felt terribly monotonous between vocals. High quality and clear in an americentric way yes but I’m not sure putting national accents in the background and scrubbing away the character of the vocalist in that way was necessarily correct either.
To be frank, it kind of reminds me of people complaining back in the vocaloid 3/4 days about accented English, which was more understandable then when there wasn’t a software developed with english-first in mind and such a wealth of english-first vocals. Now, this seems like an opportunity to meaningfully diversify the “amount of english” available.
Hoping the aforementioned slider will have something to satisfy everyone though.
I am personally pleased with how much Japanese has improved on SV2. The new model is noisier, yes, but potentially more expressive? Will be waiting to try it before making a judgement call.
My one for sure gripe that I may or may not have mentioned already is that the super dramatic autotune vibrato is even more distinct in SV2 which I don’t personally like. I wonder if more “theatric” singing presentations were rated as “more realistic” during user testing. It would explain a lot.
I think I prefer this SV2 that seems to have a distinct design philosophy and pretty distinct backend from SV1 to the alternative future which would be an “SV2” that is more of a direct upgrade to SV1.