CeVIO Kafu announced for CeVIO AI (Early 2021)

[VocAddict]

I do find it interesting that this is the first singing synth that has this problem. We have celebrity voice banks in the past with bigger names I wonder why this one is having this dilemma

As a VTuber, having a synth that is too similar to the original can cause issues because then you have two virtual representations of the same voice competing in the same landscape compared to what Vocaloid et al. is with a human vs a synth. Kaf vs KAFU will cause more friction between fans compared to Lia vs IA as the latter has distinct areas of focus.

However, in the same way that we haven't had issues with Gumi, Gackpo, Sachiko, etc. in regards to the Vocaloid usurping the VP (though during development, that was a comment raised every single time), it's only a matter of time that fans of Kaf just accept that KAFU is an extension of Kaf for others to use and it doesn't replace her.

Wow, I was right. Almost 6,000 people DID vote and TYPE A won. Kaf will take the opinions into account and be the one to make the final decision, which will be announced tomorrow.

The comments on the survey tweet really made it looked like B would win but it appears that people wants to have a voice like Kaf to use and not something to preserve Kaf's individuality. Bless all of them.

The weird part is that there is way more people voting for type C than I thought would be lol

 

[___]

If Kaf will give type A a pass...in the end I feel like it's the same voice but targeting different audiences. I feel like they're upset now but eventually they won't give Kafu a second thought simply because she won't be brought up in their circles after this settles down.

 

[uncreepy]

People celebrated too soon and assumed Kaf would pick TYPE B, but she didn't.

I translated the letter she wrote explaining her decision.

I imagine it must've been a difficult decision few people have experienced (making a synth who will go through "idol activities" just like you). And she must've felt strongly about it, since thousands of people couldn't change her mind. We will have to accept that KAFU is her own entity and why Kaf doesn't want them to sound exactly alike.

I honestly wished they hadn't shown us the results of the poll in the form of a pie chart. But @CdrSonan pointed out people probably would've gotten mad at other fans, because they would've assumed they voted for TYPE B. So it seems like a lose-lose situation for staff.

Edit: I assume Kaf also has plans to duet with KAFU, so having identical voices might make that difficult?

 

[lIlI]

A dramatic twist!

This totally reminds me of how, when you submit 3 options to your boss, they always pick the one you don't want. You gotta be careful when letting people have a choice, Kaf. Shenanigans aside, at least Type B's editing appears easy for fans to undo if they wish.

 

[Exemplar]

 

[___]

I'm not against another cutesy voice as much as I am displeased with...well...the lack of direction? To me it just seems as if they weren't even expecting Kafu to come out this realistic and then went "...Oh shit" and pitched her up. I guess what I don't like is that type B comes across as afterthought more than anything. I wouldn't be as/at all displeased if they actually asked Kaf herself to put on a cuter tone for the vb.

And I still stand by that opinion.

Before I continue I wanna state that I respect the sentiment that Kafu and Kaf should both stand on their own and the very thought of that is not what I'm bothered by at all, I just think it's disingenuous in this context, as in choosing AI synth that's more guranteed to faithfully and realistically recreate a vocal, use white Kaf reskin as the mascot, basically have the same name just slightly differently pronounced. Everything indicated that Kafu WAS supposed to be faithful recreation and the decision whether she should or shouldn't be should've been done during production and shouldn't have been a last minute decision as ultimately this isn't offering people an alternate version of Kaf's voice but shittier version.

 

[uncreepy]

I just hope that when we hear more demos, maybe people will be able to warm up a little to the voice. It would be a shame if everyone hated it/turned their back on it completely because we had what we wanted "taken" from us.

We only collectively heard 50 seconds of the KAFU voice. Maybe there are some nice things that can be done in CeVIO AI with it.

I think if many producers decide to have resentment about KAFU, Kaf's team will most likely still use/promote KAFUthemselves, because Vocaloid producers help make the original songs Kaf sings. So Kaf's huge fanbase will probably like/accept KAFU even if synth users hate her.

 

[peaches2217]

It’s not the voice people are so miffed about (though it IS just a poorly-pitched-up Kaf from what we’ve heard and I do hope the final product has more thought put into it, though being so close to release I’m not optimistic). Like, her team came out and said “Hey, we’ve heard your feedback, and we wanna know what YOU would prefer!” Then, after receiving our feedback, they glanced at it, tossed it in the shredder, and proceeded as if nothing had happened.

I respect that Kaf herself should have the final say, but what was the point in giving us a poll to express our opinions when none of them will be taken into account in the end? And even if Kaf’s fan base is what’s being taken into account, you have to remember: this is a vocalsynth. Shouldn’t synth users have at least part of a say in a product they might use?

It’s not a matter of not being given what we wanted, it’s a matter of being indirectly told that our feedback doesn’t particularly matter and we ultimately never had a say in it, after being led to believe that they would consider listening to what we had to say. That’s just... it’s gross. It’s unprofessional.

 

[uncreepy]

It’s not the voice people are so miffed about (though it IS just a poorly-pitched-up Kaf from what we’ve heard and I do hope the final product has more thought put into it, though being so close to release I’m not optimistic). Like, her team came out and said “Hey, we’ve heard your feedback, and we wanna know what YOU would prefer!” Then, after receiving our feedback, they glanced at it, tossed it in the shredder, and proceeded as if nothing had happened.

I respect that Kaf herself should have the final say, but what was the point in giving us a poll to express our opinions when none of them will be taken into account in the end? And even if Kaf’s fan base is what’s being taken into account, you have to remember: this is a vocalsynth. Shouldn’t synth users have at least part of a say in a product they might use?

The top of the survey literally warned us that Kaf would make the final decision ("Using everyone's opinions as a reference, we will decide on the voice with Kaf herself!"). They didn't promise anything. Everyone had their own opinion, but Kaf did, too. Kaf's letter today said she was grateful for everyone's feedback and read it all. The way it was written made it sound like she struggled to convince herself that using TYPE A was a good decision and she was sorry if it hurt anyone. I don't think anyone expected there to be so much feedback. And doesn't it make sense that her team and especially her would have plans for Kaf and know what the consequences of picking each option would be?

Kaf's team also posted that everyone's feedback was not meaningless. The decision was controversial even in Japan, but it seems most people are willing to accept it and trust them.

Sources:



Kaf's team implies that there is a possibility TYPE A could be used, but it is for now being set aside.

 

[Blue Of Mind]

I've already said this in the Discord, but while I am disappointed they won't be using Type A after all, I won't criticise Kaf's decision because it's not my call to make. It's a VB being made with her voice, not mine. Saying that, I do agree that Kafu's announcement and promotion so far have been a hot mess, and I'm worried how it will affect her sales once she's released. I will also remain forever confused as to why they just didn't do append banks for Kafu, because that would have pleased everyone.

On the other hand, I am looking forward to well-tuned songs featuring Kafu Type B. I'm pretty sure some producers have already formed ideas about what to do with the cuter bank.

 

[uncreepy]

Everyone only knew about KAFU for a little bit over a week and they got really possessive about it. Kaf's team has been working on her for much longer than that. I assume Kaf had plenty of time to think about KAFU's future and her own future. I do not think she would have maliciously picked TYPE B. TYPE A was obviously winning. Maybe she hoped reading opinions would convince her to change her gut feeling of wanting to change KAFU's voice to be slightly different from her own, but it obviously didn't change her uneasiness. I feel like even though it was disappointing for people, don't you think it is kind of bold to go against thousands of people's vote? Surely she had a good reason. Maybe people who voted for TEAM B had good arguments for having KAFU be her own entity, too.

 

[peaches2217]

My question is: if her team’s been working on her for so long, how many opportunities did they have to correct the issue of her sounding too similar to Kaf? It’s not like they got all of the recordings, polished them, rendered them, then whoops! Last minute surprise, it’s too similar! They had at least a handful of opportunities to change the direction of Kafu’s end output, but they didn’t, and they rectified it by... slapping a low-gen factor on the finished bank? If they intend to have her released by year’s end, they don’t have time to redo the recordings; what we’ll get is something maybe more polished, but similarly low-quality to what we heard in her demo. Or, far less likely but far worse, they’ll rush to make new recordings and remake the bank, and we’ll end up with Yuudai 2.0.

 

[Prism]

If it's anything like other ai voice banks they probably didn't know what it'll sound like. If it's anything like neutrino they take about 60 minutes of singing maybe more because it's a professional product and label it with the phonetic data and pitch and throw it into an ai training software and the voice bank spit back out. It's not based upon samples where you can have a pitch done and test it like vocaloid or synth v. What they can do is edit the original data and retrain it and it might sound a lot better think what crypton does. But it might cost more money to retrain so it might be a last minute slap on higher forment because they can't do anything else.

 

[peaches2217]

Even if it works differently, that doesn’t really mean they didn’t have a chance to change it until now; if they’ve been at work in this bank for some time, even if the method is different than what’s traditionally done for, say, Vocaloid, they still should have had some idea of what she’d sound like LONG before now. I just wonder why they’re just now doing anything about it.

 

[Stardust]

I know AI synths most likely do not work the same as Utau, but, I wonder why they did not have Kaf record a bit, then put it through the engine to hear how it sounds?
The recordings can sound very different then how Utau processes it, so if they were concerned about this, they should have made sure that the AI would not replicate her voice perfectly by asking her to do a different voice or by editing it after processing it.

Even if it works differently, that doesn’t really mean they didn’t have a chance to change it until now; if they’ve been at work in this bank for some time, even if the method is different than what’s traditionally done for, say, Vocaloid, they still should have had some idea of what she’d sound like LONG before now. I just wonder why they’re just now doing anything about it.

I suspect the backlash from Kaf's fans could be the reason why. Perhaps they saw that not too many worried(Though I am sure some worried) about Gackpo, Sachiko, etc replacing their VPs, so they presumed Kaf's fans would do the same?

edit: grammar

 

[Prism]

They might not have the bank until now. Cevio is known for doing most of the work on cevio banks so they are probably in crunch trying to train all these banks for these important companies and release Cevio ai so they just might be now getting her to her company.

 

[Twillby]

Making the poll results public was a huge mistake. They should have let only Kaf see them and then announced her decision. That way we could have assumed B won and it probably wouldn't sting as much. I don't know why I felt so bad reading the news because a tiny part of me expected this and I even said I cared about Kaf's feelings on my survey. I think the poll results let me hope a little too much X'P

Trying to spin it in my brain as, "I'll take the fact A is possible and look forward to the chance of non-celebrity CeVIO AI banks that might be allowed to sound like their voicer" but I'm having a hard time when every recent announcement has been somehow disappointing. I still want to believe in CeVIO because the snippets of A and edited B sound great but things keep testing my faith.

 

[peaches2217]

I’m with you there. I feel like they made us feel our voices mattered more than they ultimately did, and by doing so... it just feels insulting. I wouldn’t feel this strongly about it at all had they not posted the results publicly.

 

[WyndReed]

I don’t mind the concept behind TYPE B, but the distortion from being pitched up bothers me. I wish that the developers had gotten Kaf to voice act some instead of editing the samples.

 

[___]

I don’t mind the concept behind TYPE B, but the distortion from being pitched up bothers me. I wish that the developers had gotten Kaf to voice act some instead of editing the samples.

You said what I've been trying to say but in less complicated and smarter manner thank youuu