PUZZLE CVVC reclist

[Borginia]

Hello!
I've written already in the utau discussion thread about this idea, and now I have some small results!
for all people who don't know, here is the post I did a day ago.

Also, I got kinda now a ideo to create a c-cv-vv-vc voicebank (its not really that complicated thou)

its a small idea of recording a complete cv voicebank + solo consonant sounds + a vcv set just of vowels and some chain sounds ( as example npapepipopupa)

as example, the vv set can be used for soft overlap between vowels and the chain set can be used to imitate the natural vowel fadeout between vowel and consonant. (as example in "a ka", the "a" before the " " has the natural fadeout for "ka" so you would need to oto the sound "a-k".)

basically, its kinda like a vcv, just in thousand parts. (and not so heavy)

pros:
with much work, people can get control over the voicebank and can immitate vcv and human pronunication.
Is also less heavier than vcv voicebanks

cons:
need muuuuuch effort in editing a ust. (thats why the cv set is added, so if people don't want to do a c-cv-vv-vc ust, they can just go with cv or cv-vc)

and one of my results is now here!
I did this with the uncompleted rec-list of the voicebank

http://soundcloud.com/unbekannt-p/c-cv-vv-vc-voicebank-test-sky

pure vocal:

http://soundcloud.com/unbekannt-p/pure-vocal-c-cv-vv-vc

I know, the voice don't fit to the song, but I tried my best (because its a semi fast song, and it seems to work pretty fine)
(also I recorded that voicebank at night, don't wanted to get angry neighbours omg)
I used for this cover just the natural vowel fadeouts for sharp consonants (and for m and n) and mostly those sounds very fluid, in my opinion a bit tweaky, but more fluid as a cv voicebank and kinda like a vcv.

the only problems i heard out were the sounds that had no natural vowel fadeout sets (g,r,h,d,b) but I've added now for every sound one of chain recordings, so it would be possible to get natural fade out sounds for all consonants. the only exception would be "h", since its hard for samplers to stretch that consonant. For "h" I added a special vcv set so it gets at least a smooth overlap.

now after rethinking about the voicebank size, it could go about 50 mb - 70 mb (depends on quality and recorded sample size)
though, its possible to fine tune the natural vowel fadeouts for more realism and that was actually the main aim of the voicebanks purpose lol.

what do you think about it? anyone who want to help me? the reclist is already about 95% finished, so it might be probably soon up if someone want to test record for me, i would be very happy :'D

sincere, Emoto/tsundere-saurus-rex

EDIT: The reclist is finished!
And my UTAU rEmxAN got a PUZZLE cvvc voicebank as well, so people can reference from it!
https://soundcloud.com/unbekannt-p/ballade-01-remxan-puzzle-cvvc
the download for the voicebank and reclist is in the link!

Thanks Cdra for your help! I am glad that you had that good idea to make also vcv sets for the y sounds ahah

I wish you all a good day!
Sincere Borginia/Emoto/Tsundere-saurus-rex

 

[Fawkesy]

RE: A possible new kind of voicebanks?

This seems like a really cool concept, and I'd love to test record for you~ :D
Sorry I don't really have anything else to say, lol. x.x

 

[Borginia]

RE: A possible new kind of voicebanks?

This seems like a really cool concept, and I'd love to test record for you~ :D
Sorry I don't really have anything else to say, lol. x.x

oh really? it would be awesome omg
Do you have recorded any utaus so I can hear your voice? I have some people who want to test record, so I need to choose some certain people. its also recommended that the recorder already knows how to use cv vc or vcv or both haha

 

[Angel]

RE: A possible new kind of voicebanks?

I would like to test out this new type of recording style.
It's a nice change from VCV~ =w=

Anyway, my UTAU Suika has quite a sharp tone to her, so it'll be nice to see how this can smooth her voice a bit.

Sample: [soundcloud]https://soundcloud.com/angelcookie/suika-irie-piece-of-junk-world[/soundcloud]

And yes, I'm able to use CV, CV VC and VCV.

 

[Fawkesy]

RE: A possible new kind of voicebanks?

oh really? it would be awesome omg
Do you have recorded any utaus so I can hear your voice? I have some people who want to test record, so I need to choose some certain people. its also recommended that the recorder already knows how to use cv vc or vcv or both haha

I know how to use vcv~ If you need more samples, just check my YouTube~
http://www.youtube.com/watch?v=059VUwGnsc8
http://www.youtube.com/watch?v=pe16F1WZz28

 

[kirrinx]

RE: A possible new kind of voicebanks?

Cdra's done something very similar to this already ouo;; and 40-60mbs is pretty big for a bank , her Utauloids bank size is 11mbs.

[video=youtube]http://www.youtube.com/watch?v=SR0aGZr0XVc[/video]

Halt has a new CV voicebank! It's actually useable, woohoo! I used it for this just so you can hear it. You can download it here: http://www.mediafire.com/download.php...
*This is a rentan-encoded CV bank, so it has some aliasing/otoing for a little VCV, including full "VV" (vowel blends), and mostly-complete CV VC. But it works fine as a CV bank! This particular cover uses all three styles to get the smoothest effect possible. Also it has ending breaths, which I don't use anymore because I prefer vowel enders. ouo

Technically this has been done for a while but I have a big backlog of covers and stuff oop

Enjoy~

THe only difference in her bank is that it doesn't contain any " C" samples oto'd into it or have a complete " VC" endings but it does contain end breaths

and like you've stated for your method this onecan be used as a CV bank alone or tweaked to sound more realistic almost vcv like as well o 5o;;

 

[Umbreon126]

[retracted]

 

[Halo]

RE: A possible new kind of voicebanks?

I apologize, but isn't this just CV VC? They're capable of everything you describe and I've never seen one without VVs.
But that said I don't know what the list looks like so I wouldn't have a clue. The trick could be in the recording. May we take a look at a sample UST or list?
EDIT: Wait, re-read the starting post. It looks like CV VC to me.

Also, @Umbreon126
"a k" and such need to be oto'd pretty specifically, so they don't double up and sound as "ak ka"
It's simply that the vowel tapers off in a different way before "k", so it's still needed. Call it "a k" but it should really be "a [space before k]".
I just spent like 24 hours straight working with a bank like this so hey

 

[Cdra]

RE: A possible new kind of voicebanks?

This is basically CV VC Japanese, isn't it?

http://www.youtube.com/watch?v=qhIY_Cpmjks
(this is a 5-pitch CV VC Japanese @ ~100M

http://www.youtube.com/watch?v=vkWUqmDpQAE
(this is a 13-pitch CV VC Japanese @ ~240M

I make these types of banks instead of normal CV banks because they're fully operational as CV, but also really flexible with the CV VC smoothness. They don't have true VC endings (because those aren't necessary for Japanese, though if you wanted to do say Spanish or Latin you'd need them) or single consonants (but you could oto those out if you wanted them for some reason)... and of course, they have VV strings. I do think adding VCV strings for h/w/y could help a lot, but as it is, these types of banks are pretty nice.

Does your list look anything like this?
b_mbabibubeboba
by_byebyabyubyo
ch_nchechachichuchochi
and so forth?

 

[Borginia]

RE: A possible new kind of voicebanks?

This is basically CV VC Japanese, isn't it?

http://www.youtube.com/watch?v=qhIY_Cpmjks
(this is a 5-pitch CV VC Japanese @ ~100M

http://www.youtube.com/watch?v=vkWUqmDpQAE
(this is a 13-pitch CV VC Japanese @ ~240M

I make these types of banks instead of normal CV banks because they're fully operational as CV, but also really flexible with the CV VC smoothness. They don't have true VC endings (because those aren't necessary for Japanese, though if you wanted to do say Spanish or Latin you'd need them) or single consonants (but you could oto those out if you wanted them for some reason)... and of course, they have VV strings. I do think adding VCV strings for h/w/y could help a lot, but as it is, these types of banks are pretty nice.

Does your list look anything like this?
b_mbabibubeboba
by_byebyabyubyo
ch_nchechachichuchochi
and so forth?

Haha, I already looked into a cv-vc voicebank, and my voicebank actually looks a bit different.

instead of recording everything at one try (like the most cv-vc voicebanks seems to do) I record a 2-4 seccond long sample of a cv sound.

for the natural vowel fadeout, I written a other kind of recording list that goes like this:
んじゃじぇじじょじゅじゃ
n ja je ji jo ju ja

so I can oto for every sound a natural vowel fadeout uvu

well @halo: to be honest, I combined different styles of voicebanks for this, thats why I call it "c-cv-vv-vc" voicebank, because it can be used as cv, yet again as cv-vc or c-cv and much more xD the main purpose of creating this voicebank is, to extend cv voicebanks with those extra chain recordings. also, the one I did is about 50 mb, its maybe because I recorded a bit too much silence and too slow pronunication haha.

@cdra: I think I gonna send you the reclist, since you write own reclists and I am intrested i your honest opinion uvu

 

[KitWistful]

RE: A possible new kind of voicebanks?

This is basically CV VC Japanese, isn't it?

http://www.youtube.com/watch?v=qhIY_Cpmjks
(this is a 5-pitch CV VC Japanese @ ~100M

http://www.youtube.com/watch?v=vkWUqmDpQAE
(this is a 13-pitch CV VC Japanese @ ~240M

I make these types of banks instead of normal CV banks because they're fully operational as CV, but also really flexible with the CV VC smoothness. They don't have true VC endings (because those aren't necessary for Japanese, though if you wanted to do say Spanish or Latin you'd need them) or single consonants (but you could oto those out if you wanted them for some reason)... and of course, they have VV strings. I do think adding VCV strings for h/w/y could help a lot, but as it is, these types of banks are pretty nice.

Does your list look anything like this?
b_mbabibubeboba
by_byebyabyubyo
ch_nchechachichuchochi
and so forth?

Haha, I already looked into a cv-vc voicebank, and my voicebank actually looks a bit different.

instead of recording everything at one try (like the most cv-vc voicebanks seems to do) I record a 2-4 seccond long sample of a cv sound.

for the natural vowel fadeout, I written a other kind of recording list that goes like this:
んじゃじぇじじょじゅじゃ
n ja je ji jo ju ja

so I can oto for every sound a natural vowel fadeout uvu

well @halo: to be honest, I combined different styles of voicebanks for this, thats why I call it "c-cv-vv-vc" voicebank, because it can be used as cv, yet again as cv-vc or c-cv and much more xD the main purpose of creating this voicebank is, to extend cv voicebanks with those extra chain recordings. also, the one I did is about 50 mb, its maybe because I recorded a bit too much silence and too slow pronunication haha.

@cdra: I think I gonna send you the reclist, since you write own reclists and I am intrested i your honest opinion uvu

Looks exactly the same to me, but I don't think it matters. Don't get excited about discovering a new CCVCCCVVVCCVVVVFJFXXV, the exciting thing here is the goal of making a beautiful and intuitive voicebank that will have all the cool stuff a user could need or want 8D That, finding easier/more powerful/more comfortable reclists, since there's things like the difference between 5 mora and 7 mora that make a difference to voicers, even if they're both VCV. A possible new kind of voicebank ain't worth worrying about at this juncture.

But a usable voicebank is beautiful, and if you have all these inclusions then I think that'll make a big difference : D

 

[Borginia]

RE: A possible new kind of voicebanks?

Looks exactly the same to me, but I don't think it matters. Don't get excited about discovering a new CCVCCCVVVCCVVVVFJFXXV, the exciting thing here is the goal of making a beautiful and intuitive voicebank that will have all the cool stuff a user could need or want 8D That, finding easier/more powerful/more comfortable reclists, since there's things like the difference between 5 mora and 7 mora that make a difference to voicers, even if they're both VCV. A possible new kind of voicebank ain't worth worrying about at this juncture.

But a usable voicebank is beautiful, and if you have all these inclusions then I think that'll make a big difference : D

aaaa someone who really thinks my voicebank can be useful aaaa *A*
thank you really much! I finished the oto.ini as well and well, I don't know, I just leave the links here for anyone who wants to try them out uvu

romaji puzzle cvvc reclist
hiragana puzzle cvvc reclist
if you guys find something to make this voicebank easier, please tell me, okay? I really want to keep improving this different recording style ahah.

the pros i found in this voicebank type (cvvc would have also those pros thou)
-they can immitate japanese pronunication better, (like instead of [te][te] for "tette" you can do it like [te][e t] [te] and can immitate that special pronunication!)
-smoother change between tones or appends (as [a w][a k w][ka p][a t p][pa w] w=whisper, p=power, in vcv there is the danger that the appended vowels overlap and it might sound very strange)

well thats every pro i found
oh and that a puzzle cvvc voicebank has the weight of 25-35 mb, depends on lenght, quality and microphone ahah

sincere borginia/emoto

 

[Kreic]

RE: A possible new kind of voicebanks?

hmmm let me see if I understood the concept. you mean making a full VCV, CV and VC just by otoing some sounds like "npapepipopupa" using their natural vowel-to-consonant fade out? if that's the case, It works. I already did that experiment a while ago and made a VCV, CV, VC UTAU. It's an original reclist I made while trying to make my own VCV utau...but I thought everybody else used their VCV voicebanks to create a "lite" CV version of their utaus by otoing the original VCV recordings. Here's a demo

[video=youtube]https://www.youtube.com/watch?v=pB40SfSpR8c[/video]

 

[Borginia]

@Kreic: Indeed! well, cvvc japanese use the same concept, thou there is a otoing method I use that makes this kind of voicebank kinda different from the others ahh

https://soundcloud.com/unbekannt-p/ballade-01-remxan-puzzle-cvvc
Here you can download the reclist and my voicebank I've did with the cvvc PUZZLE reclist!

its also easy to convert a ust to fit a puzzle cvvc voicebank!

just follow those steps:

(0.) If your ust is in romaji, convert it to hiragana with the romaji to hiragana converter!)

1.) Use the auto vcv plugin to make the vcv like additions work (like [a a] [i ha] [u ya] and etc] (PS: I guess every utau user has that plugin, so I think it

2.) Use the Autocvvc plugin for namine ver.1.0 (downloadable here) to convert the ust for the puzzle cvvc.
try to copy those settings:

also, that thing with the "60".. it depends the lenght of the vc endings!
60 is for normal and fast songs, while 120 is perfect for normal and slow songs.

3.) after you converted the ust, you can start tuning! (dont tune before you convert, or else the convertion might delete some of your tuning..)

4.)Also check if some vc endings are fast enough to play, because some of them (like t, p and etc sharp consonant) might be too slow to be played, then you need to delete them. the [R] would make a "fake" vowel fadeout for the note before and the difference is not easy to hear out. for a normal-fast song, the ust would look like this:

5.) after editing and tuning the ust, render the ust and then click "crossfade optimization"

it will make the crossfade between the vcv,vv, cv and vc really smooth and softer!
not many people seem to know about it what is kinda sad, but that tool is seriously the best thing in utau

and this was it actually ahah.
thank you for reading

Sincere Emoto/Borginia/Tsundere-saurus-rex

 

[Borginia]

I am sorry for bumping! I just wanted to say that I've finished the puzzle cvvc reclist!
Its downloadable here
Thanks to cdra I improved the reclist!
Updates in the reclist:

• VCV strings are in 4Mora or shorter for easy to pronounce.
• VCV/VV strings for h, y, and w's for smooth overlap.
• Pre-aliased oto.ini! (edit the oto in setparam as you want)
• has a tutorial bgm like vcv!
• Demos can be heard on my SC channel!
• First complete puzzle reclist will be released soon! (rEmxAN 1NEN)
• ?????????
• No plugin for complete convertion. you need to use two seperate plugins to make the voicebank work

Sincere Emoto/Borginia

(I will try to be more active here uvu)