CeVIO Kafu announced for CeVIO AI (Early 2021)

[uncreepy]

The details aren't public yet other than that it's coming in winter 2020/being a singing synth voice, but a virtual singer named Kafu (they write it as Kaf, though) (花譜) is collaborating with CeVIO. She started VTuber activities when she was only 13 years old (she is now 16) and her popularity quickly grew. This is what she normally looks/sounds like:



Her design for CeVIO has changed to have white hair and her name kanji changed to be 可不:

 

[msmsmsm]

Pretty! Going with CeVIO AI is definitely a good call for Kaf since she has such an interesting timbre and delivery, and other synths would probably do it an injustice. I'm very excited to see how this will turn out.

 

[uncreepy]

Translation of the text on the announcement image:

音楽的同位体
Musical isotope (What the heck does that even mean?)

花譜から可不へ
From Kaf to Kafu

聴くから創るへ
From listening to production

The artist for the announcement is PALOW, this is their Twitter: https://twitter.com/PALOW_
It looks like they have done album art for Kaf in the past. They have also drawn fan art/album art for Vocaloid before (Miku related).

Edit: @CdrSonan wrote a simple explanation for the "musical isotope" thing:

 

[First Sound Future]

Wasn't really interested when I saw that she's a VTuber at first but seeing her channel, she seems to take more of a "mysterious" approach which shows in the Cevio design as well. Her voice is pretty unique tbh I wonder how is it actually going to sound in the engine. Also, there seem to be more VTubers from the same company as her, so there might be more Cevio products from them in the future, hopefully. One of them even sings English songs apparently? Here is the page on the Virtual YouTuber Wiki

 

[Prism]

Do we have any examples of cevio ai singing yet?

 

[uncreepy]

The reason they probably picked Kaf is because she is known for having a whispery yet wavering powerful voice, so a lot of Japanese users are interested in being able to create with that type of sound. So at least it's a new voice type compared to the same old usual.

Do we have any examples of cevio ai singing yet?

Unfortunately, nothing yet.

 

[Stardust]

Judging by the text in the image, it seems we have a fairly concrete date for CeVIO AI now. Instead of just Q3(or was it Q4? I am misremembering), we now know it's Winter 2020, which is most likely December. Unless they count November as part of Winter, but I do not see people usually do that.
With a release date so close, hopefully we will have singing examples soon.

Somewhat unrelated, but I am surprised it took so long for a Vtuber to join the Vocal Synth scene. I hope this means Aoi will provide a voicebank, her voice is very unique. Mori would be wonderful as well, but she is too new, and her genre is rap, which most synths cannot replicate well yet unless you are very skilled like Mitchie M.

 

[uncreepy]

A YouTube channel has been made for Kafu.

Announcement:

(This is not Kaf's normal Twitter account, this is a new one for Kafu. Also, right now it gained 2,746 followers in like 2 hours.)

The only video on the YouTube channel:

Singing starts at 0:32

I am not sure what I think, she kind of reminds me of Aoki Lapis (texture) + Tohoku Kiritan (pitch).
She's kind of hard to understand in this clip, too. (Try to listen for the vowels and stuff before reading the lyrics I have under this sentence.)

Here are the lyrics courtesy of Clover (クロバー) in the comment section:
許せる気がした / Yuruseru ki ga shita [Fades in so you can't hear the first few words]
さよならだよ全部 / Sayonara da yo zenbu
忘れてしまえ / Wasurete-shimae
日暮れの匂いと共に霞む言葉 / Hi-gure no nioi to tomo ni kasumu kotoba
前だけを見なよ / Mae dake o mi na yo
私を置いて / Watashi o oite
君のいない夏へ / Kimi no inai natsu e
君のいない空へ / Kimi no inai sora e

Edit: Here is the song being sung by Kaf so you can compare (timestamped), thanks @Kazumi :

 

[Blue Of Mind]

I am not sure what I think, she kind of reminds me of Aoki Lapis (texture) + Tohoku Kiritan (pitch).

I'm kinda disappointed with the voice so far, because when I saw the character design, I thought a deeper tone would have been the most appropriate (I'm not the least bit familiar with Kaf the VTuber at all).

 

[___]

I will have to listen to more examples to make definitive opinion but I'm nooooot sure whether Kaf's unique vocal traits carried over into the synthethized voice well, her voice is more complex in texture and delivery but here the engine seemed to turn those characteristics into..like...general nasally sound?

And while Kaf's capable of delivering this cutesy tone she doesn't consistently sing in this tone either...but that's all this vb seems to be? Again, will have to listen to more examples once they come out. I'm just hoping this was just them getting carried away with the gender parameter/whatever the Cevio equivalent of that is.

 

[xuu]

She sounds pretty funky, I like it. I think there's a bit of filtering going on with the voice? Looking forward to hearing a clearer sample soon, but I like what I hear so far.

 

[uncreepy]

(@Blue Of Mind In case you missed it since I just edited the post, I have the timestamped version of real Kaf singing the same part of the demo song added at the bottom of my previous comment)

Well, she is a teenager, so it's not like her voice is going to be that deep. But anyway, I feel like they either increased the gender parameter a bit for this demo so she sounds a little high pitched compared to real Kaf, or maybe this is what you get by default. Not sure if CeVIO AI lets you edit parameters, but you could edit the pitch a little after exporting.


Amano extracted her singing from the BGM, so the quality got a little distorted, but he tried to make it sound more like Kaf by editing the pitch a bit.

We will know better if we get to hear more demos, I think.

 

[Kazumi]

Hi y'all! I pitched the demo of KAFU AI down by 2 semitones, and attached the result. I'm... amazed to say the least at how much she sounds like Kaf. Well done CEVIO AI, well done.

https://clyp.it/h5tjzfap

 

[___]

Ah! Thats much better! Really hoping they just edited the pitch then and this is what default Kafu will sound like:o

 

[lIlI]

She sounds like a high quality Gachapoid and I don't know what to do with this information

 

[peaches2217]

I really, really wanted to like her... I dunno. So far I just... don’t. I do hope official demos do her more justice, but from this, I just can’t vibe.

 

[Rylitah]

I really can't think of a reason they'd do this. People already liked her original voice. Pitching her up then is....??

Samples like this to me seem so disingenuous; this sound is what people who aren't familiar with vocal synths are gonna expect right out of the box. Unless that really is what she sounds like on default, which I find.. hard to believe, especially since she sounds just like her VP when pitched down.

Man, I dunno what to think, haha. I like her normal tone but everything about new CeVIO lately has been so confusing and wild and none of this is really helping me warm up to it.

 

[VocAddict]

Must say I'm disappointed in that demo, really wasn't expecting that at all but I have a feeling this is not how she sounds in her default state. If that's the case then.... I guess this will be the first white-hairded character I'll be disliking right off the bat.

I know it's probably not right to compare it to the technical demo they did back in 2018 where they did comparisons between the the old and new tech with IA so I have hopes that Kafu would end up sounding much closer to Kaf's at the end (please please be a badly tuned demo), unless this is the developer's intention then ¯\_(ツ)_/¯

 

[mobius017]

Amano extracted her singing from the BGM, so the quality got a little distorted, but he tried to make it sound more like Kaf by editing the pitch a bit.

Hi y'all! I pitched the demo of KAFU AI down by 2 semitones, and attached the result. I'm... amazed to say the least at how much she sounds like Kaf. Well done CEVIO AI, well done.

https://clyp.it/h5tjzfap

If the only difference between those samples is a 2-semitone pitch shift, then I think the distortions in the first one really make it more difficult to judge that first sample fairly. It's almost a subconscious effect, but it's hard to shake how bad that first one sounds because of the sound being messed up. On the other hand, that second sample, if it's really just shifted 2 semitones down, sounds really good.

If they shifted the final vb, maybe they did it to make it sound more moe or something? (Not sure if it really gives me that vibe--kinda yes, kinda no, the more I listen--but maybe that could have been the intention.)

 

[uncreepy]

Must say I'm disappointed in that demo, really wasn't expecting that at all but I have a feeling this is not how she sounds in her default state. If that's the case then.... I guess this will be the first white-hairded character I'll be disliking right off the bat.

I know it's probably not right to compare it to the technical demo they did back in 2018 where they did comparisons between the the old and new tech with IA so I have hopes that Kafu would end up sounding much closer to Kaf's at the end (please please be a badly tuned demo), unless this is the developer's intention then ¯\_(ツ)_/¯

The problem with those demos from 2018 is that the engine wasn't optimized for being able to sell to normal consumers. It took 5 hours to render a 5 min song. So they had to compromise/balance it to get it to a point the software was better than CS7 but actually usable for most people.

I wish we knew for sure if we could even TUNE in CeVIO AI. Because they had to edit the demos they showed with Melodyne, so that sorta sounds like... no parameters to me??

They really are gonna make us wait until the last second to see the actual software, aren't they? *sigh*

I wish we knew more.