What I don't understand is why they would do CeVIO if it costs more than making a Vocaloid. It's been touched on here and on the Wikia, but I still just don't get it. All Ia's other voicebanks have been for Vocaloid, so it's a real bummer that it's now on CeVIO. I'd love to have bought her Vocaloid bank, because I just feel like that would've been better.
Esp. considering CeVIO is the whole "jack of all trades, master of none" in that it can do more, but it's a lower quality result.
Even if a C voice is harder to make than a V one, it doesn't matter for 1stPlace, since the CeVIO company (techno-sp.) would handle that. The whole voice (ONE / IA) is made by CeVIO company, they just "sell" them to 1stPlace, who just need to do some tweaks. The VoxWave is a sad story because they have to adapt the engine for French but they don't have the knowledge (about HMM etc.) to do that at all.
I don't really agree with that "jack of all trades, master of none" words. Statistics & ML based voice synthesis is the future. Sampling and pitch shifting voice is more like an instrument - I don't mean V is useless, people like instrument-like voices as well. The only critical weakness of HMM based engines (CeVIO/Voiceroid) is the noise. Just look at the TMG parameter in C, V can't do that despite of the tricks. You might say some V songs sound more realistic than C, but that’s just because of the hard work of producers. They can do the same thing on C, too (such as 神無月P).
I agree with "Voiceroid2 talks better than C". If the voices are made in the same quality (Voiceroid voices are also very hard to make), Voiceroid2 (not V1 or V1EX+) does sound better. But the fact is that both Voiceroid and C are based on HTS, they are actually brothers - which is not true for Voiceroid and Vocaloid. The Voiceroid2 (actually it's AITalk) just made some modifications on the open source version of HTS (look at the huge difference on voicebank sizes), and C is developed with the lab which created HTS.
