nyl II currently has a lot of issues with engine noise and random artifacting in the vocals, an issue which i think has arisen simply due to the nature of how these new voices work - the same thing causing these issues is the same thing that, imo, allows the other vocals to have more of a human-like depth to them ironically. like you said too lili, some of those things that make them sound unnatural do also wrap back round to feeling natural, which is why i imagine it's an issue in the first place, and one that hasn't immediately been fixed (though knowing ES, i'm sure there'll be a solution)
it does make them a lot harder to use in my opinion, but the extensive nature of SV2's parameters also means that there's a lot that the user can manually do to work around - nay, even make soundgooodable (sorry) - these things, which i always believe is in the true spirit of vsynth
funnily enough, i've heard the opposite about solaria that her range and expression has done nothing but improve, but absolutely nyl and saros have suffered a bit of a hit in the switch-over - ironic that asterian is the most stable given how 'experimental' a low-octave ai vocalist technically is LOL.
(sorry if this is worded like crap btw. despite doing literature and linguistics this year at uni, i think my writing skills have really really deteriorated omg)
