KVR Audio

xoxos · Post by **xoxos** » Sun Feb 15, 2015 10:25 pm

Spitfire31 wrote: When will we get a raspy, sensual, fullblooded blues voice?

i dunno, bro

you keep an eye on those majors now :p

garryknight · Post by **garryknight** » Mon Feb 16, 2015 12:02 am

Spitfire31 wrote: When will we get a raspy, sensual, fullblooded blues voice?

Just keep drinking the meths and it will happen...

Seriously, though, I think they just added raspy to Vocaloid 4. But that's only available in Japan.

xplo · Post by **xplo** » Mon Feb 16, 2015 12:54 pm

teacue wrote:Feel free to report if you have time

So I've had a chance to play with the new Vocaloid4 engine and the Cyber Diva library and I'm pretty impressed with the improvements they have made. The pronunciation is much more clear and the Engrish accent is also gone, mostly.

For comparison, here's a cover of Glass Wall using Cyber Diva. I've went through and did some diva-specific fixes. Note that there's still some features untouched by this demo. Growl and crosssynth were not used yet.
I did transpose the vocals down. Mostly because it then stays within Cyva's comfort zone but also because people keep complaining about the vocaloids being only high-pitched voices.

https://soundcloud.com/xploded/glasswall-cyva
Published with permission of the original song's producer GHPZ.

You can find the original by searching for Glass Wall on YouTube if you want to a/b it and check the improvements between V3 and V4

20 years? Probably much sooner at this rate.

Spitfire31 · Post by **Spitfire31** » Mon Feb 16, 2015 1:46 pm

All right, make that 19 years…;-)

Judging by Glass Wall, this prepubescent kid might pass as a backing, buried deep down in a mix. Solo voice? Only if you want an intentional effect, smelling of vocoder.

/Joachim

teacue · Post by **teacue** » Tue Feb 17, 2015 9:54 am

xplo wrote:So I've had a chance to play with the new Vocaloid4 engine and the Cyber Diva library and I'm pretty impressed with the improvements they have made. The pronunciation is much more clear and the Engrish accent is also gone, mostly.

For comparison, here's a cover of Glass Wall using Cyber Diva. I've went through and did some diva-specific fixes. Note that there's still some features untouched by this demo. Growl and crosssynth were not used yet.
I did transpose the vocals down. Mostly because it then stays within Cyva's comfort zone but also because people keep complaining about the vocaloids being only high-pitched voices.

https://soundcloud.com/xploded/glasswall-cyva
Published with permission of the original song's producer GHPZ.

You can find the original by searching for Glass Wall on YouTube if you want to a/b it and check the improvements between V3 and V4

Thanks a lot for posting this.
I like it and it seems there is a huge improvement in V4!
But I like also your Miku version

Best regards

Jaleo · Post by **Jaleo** » Tue Feb 17, 2015 11:13 am

Vocaloid is a great technology, the synthesis quality improved a lot recently, but the editor is really limited, lacks of smart tools to take advantage of the engine capabilities.

I don't think it's a mature app for music producers actually, until new tools be developed to reduce the amount of time spent editing to achieve decent results.

IMO, at the moment the biggest problems are to properly control phoneme timing (for a clear and natural pronunciation flow) and portamento time between notes (for a better pitch entonation and performance). Both subjects are controlled mostly in the piano roll, I think it's wrong. Each needs a new editing layer where you can move start/end phoneme/portamento mark points and keep the piano roll as it is, the target notes layer.

Some samples of my work with different voices, I never use pitch correction tools:

https://soundcloud.com/blancanegra/viu-la (sung in catalan)

https://soundcloud.com/blancanegra/amag ... mavera-ona (sung in catalan)

https://soundcloud.com/blancanegra/vol-volar (let it go - Frozen, sung in catalan)

https://soundcloud.com/blancanegra/dont ... -sweet-ann

https://soundcloud.com/blancanegra/yest ... vanna-test

https://soundcloud.com/blancanegra/blue ... oloid-test

https://soundcloud.com/blancanegra/jolene-ann-test

Spitfire31 · Post by **Spitfire31** » Tue Feb 17, 2015 11:37 am

@Jaleo: Please let me get this straight – are these songs Vocaloid productions?

IF they are, then I take back whatever disparaging remarks I posted about the technology per se – apparantly, I had only heard the depressing results of inept programming…

/Joachim

Jaleo · Post by **Jaleo** » Tue Feb 17, 2015 11:52 am

Spitfire31 wrote:@Jaleo: Please let me get this straight – are these songs Vocaloid productions?

IF they are, then I take back whatever disparaging remarks I posted about the technology per se – apparantly, I had only heard the depressing results of inept programming…

/Joachim

Yes, they are!

You are right, there is a lot of bad examples, but it because the editor is very limited to work with.

If you like to see the editor in action, listen some of my jobs in my Youtube channel, you can find things like this:

https://www.youtube.com/watch?v=R2EcFqLwNQQ

wagtunes · Post by **wagtunes** » Tue Feb 17, 2015 12:01 pm

Yesterday Once More was just flat out amazing.

I'm speechless.

Okay, how long did it take you to do that minute and a half of music?

Spitfire31 · Post by **Spitfire31** » Tue Feb 17, 2015 12:02 pm

Fantastic. You're a wizard – and a stubborn, hard working wizard, I can imagine.

I'm thoroughly impressed and will enjoy your Youtube tutorials.

Now I only hope that the editing technology soon develops to a stage allowing a mediocre wizard's apprentice to inject some of the expression and sensuality that seems to be your forte.

Kind regards,

Joachim

Jaleo · Post by **Jaleo** » Tue Feb 17, 2015 12:24 pm

wagtunes wrote:Yesterday Once More was just flat out amazing.

I'm speechless.

Okay, how long did it take you to do that minute and a half of music?

Well, the question is a trap...
How long did it take to do a minute of good CGI? XD

Maybe I spent 3 hours or so in that sample. As I said, the editor is very limited. Each thing you want to do takes a lot of time because the tools (It's more like pixel art in MS paint compared to photo editing in Photoshop) New tools must be developed to reduce the editing time, because the synthesis engine is very capable.

wagtunes · Post by **wagtunes** » Tue Feb 17, 2015 12:29 pm

Jaleo wrote:
wagtunes wrote:Yesterday Once More was just flat out amazing.

I'm speechless.

Okay, how long did it take you to do that minute and a half of music?
Well, the question is a trap...
How long did it take to do a minute of good CGI? XD

Maybe I spent 3 hours or so in that sample. As I said, the editor is very limited. Each thing you want to do takes a lot of time because the tools (It's more like pixel art in MS paint compared to photo editing in Photoshop) New tools must be developed to reduce the editing time, because the synthesis engine is very capable.

I get that it's editing. My question is this. How did you learn to do the editing?

For example, take the phrase "It's yesterday once more" at the end of the chorus. How did you learn to do whatever you needed to do to make it sound like that? Was it trial and error? Certainly the editor itself, if it's so bad, can't be intuitive. So how did you figure it out? And is it something that anybody can figure out or will I be sitting here pulling my hair out of my head because I can't figure out how to get a phrase to sound the way I want it to sound?

Jaleo · Post by **Jaleo** » Tue Feb 17, 2015 12:36 pm

Spitfire31 wrote:

Fantastic. You're a wizard – and a stubborn, hard working wizard, I can imagine.

I'm thoroughly impressed and will enjoy your Youtube tutorials.

Now I only hope that the editing technology soon develops to a stage allowing a mediocre wizard's apprentice to inject some of the expression and sensuality that seems to be your forte.

Kind regards,

Joachim

Well, it's not magic, just science.

I was researching by my own (if you can call it in that way) what's worng with Vocaloid output compared with a real singing performance, taking note of the issues and figuring how to fix them just using the in editor tools.

Unfortunately, as I said, the editor is very limited and small changes that can be done in miliseconds takes a couple of wasted seconds. At the end of the day, it's a lot of hours.

Spitfire31 · Post by **Spitfire31** » Tue Feb 17, 2015 12:44 pm

Jaleo wrote:Well, it's not magic, just science.

I think that the magic lies in keeping the creative goal alive during all those tedious tweaking hours. And obviously, you know a lot about singing in order to formulate that goal in the first place.

/Joachim

teacue · Post by **teacue** » Tue Feb 17, 2015 12:54 pm

Jaleo wrote:Vocaloid is a great technology ...

Thanks a lot for posting your amazing work!
And also thanks a lot to contribute to change the mind of people sceptical to Vocaloids

On my side (very new to Vocaloid) I appreciate the realistic possibilities as well as the instrumental crazyness one can achieve with it.
I guess I even may begin to collect the Vocaloids of the first or second generation before they disapear just because of their typical non realistic sound

What is the general consensus on Vocaloid?