Other Cryptonloid voicebank updates, collabs, & concert news (crypton_wat Twitter translations)

[mobius017]

I'm going to do my best to paraphrase something I found from the machine translation I read, for the benefit of those who are interested in getting the software or in knowing when their software will be updated:


My paraphrase:
It looks like there may be a new minor update as soon as next week. Includes bug fixes, improvements to "parameter problems."

The timeframe for the original update is under reconsideration; it may be the end of June or the beginning of July. Includes improvements related to adding vibrato, some "problem improvements," and an update to the whole voice DB.

There's also an apology for this potential delay, and the promise that they're trying to proceed as quickly as possible.


There was also someone else asking if Cherry Pie would be included with NT, so that question is one that multiple people have.

 

[mobius017]

Apparently, Wat is working on something called "homophone multiple vowels" that users will be able to choose between at will. (In broader terms "homophones" are words that sound the same, like "there"/"their"/"they're." So this probably refers to something like different combinations of multiple vowels that end up sounding the same, like the long "o" sound in "road" and "toe," maybe?) I don't see the point in switching between two things that sound the same, so maybe they're just close or something.


In any case, they're apparently trying "urgently" to improve the sound quality.

 

[phantoCrystal]

Homophone multiple vowels... Well, I think that's for more natural singing I suppose? It's hard to explain, but you know when an English vocaloid sings, it almost sounds like they're singing in an accent they would talk in? Sometimes I feel like while singing, vowels are just looser and it helps for a more natural approach, so singing doesn't sound like talking... idk if that makes sense, but if that's the idea, that might be cool? Anyway, it seems that they care a LOT about what users think and are actually working hard, so... That's awesome.

 

[Twillby]

Wat told someone who replied to the tweet that it's the difference between a seiyuu saying るー and るううっ* which I think I understand on some level but it's hard to put in words. I think it's like the first one implies just plainly holding out the vowel at a steady level and the second one's got a kind of cutesy trill to it, like when someone's talking in a singsongy voice. No idea what he's planning on doing on a technical level but it sounds like it might be interesting.

* I left this in hiragana because it feels impossible to express well in romaji but my best guess would be "ruuu" and "ru~u~u~" X'D

v (I wrote this paragraph before I read wat's reply and saw it's off but I spent too much time on it to be able to bring myself to delete it rofl. TLDR I thought it was like round robin sampling at first)

See more

I have doubts this is what they’re aiming for but it kind of makes me think they might be trying to do something like this technique I’ve heard of used in other programs called round robin. When you repeat the same note on an acoustic instrument there’s still a slight difference each time you play it, so developers record the “same” sample multiple times to add a touch more realism. Key word being record though, which makes me doubt myself because they’re just editing the V2 samples instead of re-recording, right? I guess it’s possible they’re editing copies of the samples to try to create a similar effect, but I feel a little skeptical about it being as effective. Though you don’t know unless you try I guess.

 

[Prism]

I wonder if it will help with pronunciation I remember wat talking about focusing on it

 

[kozet]

The first thing that came to mind was being able to choose between different allophones of a phoneme.

 

[mobius017]

Wat has reviewed the sound quality updates to the NT database. He says they're going well and that they're working to get those updates out ASAP.

 

[Scarlet Illusion]

It’s nice to see that they seem to be actively listening to feedback. That reflects well on them, and they’re probably more inclined to listen now than back when they were still on Vocaloid because they have the freedom to release updates and patches as they please.

Kind of reminds me of how everyone reported issues with Rin and Len’s ACT1 (which is actually very good, btw. At least for Rin anyways), except now they don’t have to release a free update in secret. Now they can update as they please.

I look forward to seeing what else CFM manages to do with their newfound freedom.

 

[DefiantKitsune]

To be fair I don't think it was really a "secret"- I think Act2 were promoted pretty heavily as updates.
The Miku V4C update seems to be more of the "secret" one

 

[Prism]

Yeah act 2 was on the box. I do wonder if we are going to have an act 1 situation where they release a bank and update and there's fans of the original bank though and it becomes rare

 

[uncreepy]

Miku NT has an update (have to go to Help > Check for updates > close Piapro so only the update pop-up is left > click アップデート / update (the updater is only in Japanese, how helpful and amazing) > wait a bit, click OK > get prompted to click アップデート / update again > OK to finish (mine kind of froze up briefly while downloading the packages, but clicking the pop-up unstuck it).

初音ミクNTプロトタイプ版・不具合改善アップデートのお知らせ – SONICWIREブログ and Piapro Studio NT 3.0.0.1 アップデータを公開しました | Piapro Studio Official Website talk about it.

No, I'm not translating anything about this update. Just talking about how to install it today.

 

[uncreepy]

Going to add:
Note Setting doesn't crash anymore/actually has a pop-up for options.

They completely removed Export as V3/V4 (can only export as audio, can still import Vocaloid files).

Crypton will release updates at the end of June & July. They are calling it "beta 2" & "beta 3". lol cuz thats how beta works... just slap a number on there for endless betas and don't do it proper like alpha/beta/etc cause they have no clue how many updates they will have.

 

[uncreepy]

Spent some time trying to decide if this is vaguely worth mentioning or not, I guess I will. Got the following error message immediately after updating NT:


I couldn't do anything in the software, had to give up and close it (the pop-up has no close button other than quitting Piapro). It asked if I wanted to save when I tried to close it even though I wasn't able to make any changes to the file, though. I guess based on the heading, it's relating to them confirming our license after the update.

 

[Prism]

It is a Japanese vocal synthesizers made by a Japanese company so it's not too outrageous that they only have beta updated in Japanese. I would imagine we get a lot more English support when English voice banks come out/ finally release.

 

[uncreepy]

The actual Piapro interface is in English, though. It makes no sense to me to have the software and Sonicwire page in English but not the updater. I think it's tacky to take our money through preorders and then only provide update emails about it in Japanese (how they had 2 emails about delays in only Japanese) and have half of the interface (as in the updater) in Japanese. I can read it, but I feel sorry for people who can't. It's confusing, why would you expect that software you paid like $180ish for to have such poor notifications regarding customer service. The updater prompt also contained only the Japanese portion of the license agreement. They could have copied/pasted the English version of the agreement at the bottom and or actually translated the "Update" button. They HAVE some form of English customer support/a translator, they just aren't using them for some reason.

Edit: The day I feel sorry for a company with "only" 100ish workers will never come. I feel sorry for individual workers, not a company that knows they have an international following, yet refuses to pay a translator or 2 to translate a measly pop-up with a couple buttons on it. Sort of related, but I read the reviews on working for Crypton (2 Japanesee Glassdoor equivalents) and it stinks, apparently, and got a 3.3/5 and a 2.8/5 overall, maybe when I stop feeling salty about Crypton I will share my translations for those. : /

 

[mobius017]

Something else interesting Wat said:


Apparently, they're going to focus on improving the sound quality and naturalness of the phonemes that are already in the VB before any addition of other phonemes. I would have thought that they would have migrated all the Japanese phonemes already, but I suppose in light of this that that wasn't the case. So I guess there are possibly more phonemes to look forward to.

Also, he mentions something that seems to indicate that the VB will be made able to pick the next phoneme from some kind of branching structure. I'd guess this is possibly related to that stuff I mentioned earlier. But I don't really understand it any better than that, so I guess we'll all find out together :) .

 

[mobius017]

Apparently they're recording today. The new recordings are slated to be installed in the β3 version of the VB between mid and late July.

So I guess Saki Fujita is recording some new samples after all.

 

[phantoCrystal]

This is very interesting indeed.... But they seem to be putting a whole lot of work into making this bank sound good, and are listening closely. It also seems as if they have a LOT to add, which I'm surprised about. I thought this was gonna be one of those prototypes where they have us test it, get a teensy bit of feedback then make a small tweak before releasing her fully... Nope.

 

[Prism]

Do we have a clip of the update and the original

 

[mobius017]

Testing of yesterday's recorded audio data has been completed. Along with the January version (Of some/all of the sound files? Of the DB as a whole?) from before Corona, they're currently working on adding it into the audio DB that will be released at the end of June.

The next few upcoming versions of the prototype NT may see considerably increasing file sizes, but file size is expected to get smaller in the release version, when there will be fewer extra vowels.