Talk:Nekomura Iroha V4/@comment-4538784-20190920101759/@comment-53539-20190921085623

Vocaloids are made by recording layers of sound.

The layers come in two forms; stationary ad articulation.

Iroha was a 2-layer "stationary layer" vocal voicebank because that was the norm for V2, they almost all had 2 layers only of sound. So. No. Iroha doesn't have more layers of sound then she did in the past... But I'm referring to only the core sounds of her vocal here when I say that. This controls the main stability of the vocal and makes Iroha sound the way she is.

The extra "layers" you hear is "articulation layering", is caused by the accumulation of phonetics, diaphonetics and triphonetics. Essentially, thats the "articulation" layering. Meaning, she has more sound data then before. She has triphones in V4 she didn't have in V2 and generally more sounds.

In regards to re-recording the provider....

Iroha, Miki Yuki and Kiyoteru were released very soon after V4 started, i.e. within its first year, it is possible the 4 were actually due for V3 releases. This happens a lot, a project gets X far into the release and then the engine does its 4-year job to the next engine and projects are left over. This causes vocals once ready for the previous engine to have to be adjusted and even re-submitted into Yamaha to get the 'okay' for release. It doesn't take long, but there are 6 voicebanks Ah Software was handling with their V4 vocal releases, thats a lot of voicebanks that need to be adjusted for V4 instead of V3, and Ah Software was distributing Yukari who had 3 voicebanks they were aiding to release. Ruby, Luka Append, Kaito, Meiko, Megpoid Extend and Oliver are all vocals at least known to be impacted by engine switching.

If Iroha was recorded for V3, then not many years would have passed at all and the voice would sound basically the same enough to progress forward. Iroha "Soft" in due respects, likely is the current capabilities of her provider, as the provider has likely been trained to be better at falsetto. Falsetto is often important to those who transcend gender borders as it allows men to sound like women, and a lot of transgender women learn it to mimic how a woman sounds.And Iroha "Soft" is within the Falsetto common range expectations.

But the big deal of V3 is more triphones, so ever 2 layered stationary vocals like Miku or Iroha end up with more layers because of the added articulation caused by the increased number of sounds within their voicebanks. Thats all. They re-enforce behaviours, but they don't change how the vocal sounds overall. The majority of Vocaloid 3 vocals have 3 layered articulations, making them overall 5 layered vocals, however, studios never take articulation layers into account when they say a number of layers a Vocaloid has for some reason... We've only had a handful even mention them. Ruby, Yohioloid and Maika I think are the only 3.

So if Iroha V4 is a 5 layered sound, you know what that means. And articulation layering is hard for us to record because its... Complex. Yohioloid for example has a half-layer of articulation sound, but when talking about Ruby, Syo said he had no idea what ".5" of an articulation layer meant.

Triphonetic data can also be extracted from existing sounds so you don't need to re-record the provider at all. The only person who they had to record new samples for indefinite was Yuki, because the provider(s)(there is some hits that they used 2 providers, for her 2 layered vocal) weren't children anymore.

Considering how many studios manipulate recordings, with Zero-G being one of the few studios not doing that (save Sonika, who was their experiment to see if the way CFM does their Vocaloids actually works), a lot of studios have sound engineers. Their job is to clean up samples and fix inconsistent samples, basically ones that sound off, as well as make up samples (some sounds are confirmed as unrecoverable, meaning you have to paste two samples together to get them instead). A good sound engineer, can do a lot and even if they did re-record the provider, they can manipulate it to sound like the old provider. We're talking about studios here whose job it is is to create sounds for use in music, not just vocalists but all sorts of sounds.

As I said, what the "re-recording" can also apply to the copied files from the master recordings and in due respects, its most likely what their referring to. This counts, as I said, as a "recording". So, a "re-record" vocal here can simply refer to Ah Software having gone back and extracted the files from the master.

At the end of the day, we don't know what the case may be, we only have what Ah Software said... And they just said what they said. Sometimes the devs say stuff, and we don't get the full details, or somethings lost in translation. This was n the product page of Iroha in the V4 era, and details weren't given. We know "Soft" and Kiyoteru "Rock" were entirely new recordings altogether, but we don't know what their referring to with the "Native" vocals.

I won't speak on UTAU, because the reaction is often bad. UTAU users often jump to try and prove Vocaloid is "less realistic" and "Vocaloid is inferior" and in the end, it doesn't matter. There are differences in the way the two engines do things and I'm not going to discuss it here.

My experience is that people will just resort to cherry pick and that often what they pick isn't either as good as they make it out to be... Or... Alternatively, not a fair representation of typical results. Example, back in 2010, to make English vocaloids look bad because they had an absent of examples, people used to bring up popular Japanese Vocaloid songs... But in doing so, this wasn't a fair representation of English nor even of Japanese Vocaloid.

UTAU users are these days falling into this same trap when talking about both engines. I don't want to go there because cherry picking is pointless because people can perceive something as better or worst then it is. And I've had enough within Vocaloid itself, but recently, its starting to get bad with UTAU and Vocaloid, and thats a bad sign of the times... History for me is about to repeat itself in the worst way. When they start cherry picking, I don't respond with cherry picked Vocaloid results, because I know the UTAU cherry picking isn't a fair representation of UTAU, but neither will be a Vocaloid cherry picked result.