What is the general consensus on Vocaloid?

[raintalk]

added modulation of voiced level and a few other params... maybe still some tweaks, but i think this is more or less representative of what one can expect with this technique with a reasonable amount of work -

http://xoxos.net/temp/syng3tyg2.mp3


so... what do you say now?

I'm liking it!
That's what I think in my head when I think "singing computer"

[SODDI]

Well, I wasn't really referring to that example specifically. I mean in general, how do you know what to do with each line of music. Is it intuitive? Do you need to understand in detail how vocals are sung in real life? How hard is it translating that knowledge into the program?

Look, let me put it this way. All I know is that I open my mouth and sound comes out. I don't particularly know or care how it's created. I don't consciously think about what I'm doing when I sing.

Now, do I need to be able to do that to use this software?

That's all I'm asking.

The more you understand how vocals are sung the more easy would be. As the program only shows you high level information, you need some practice into it to figure how your changes would be rendered.

I start preparing a midi track of the sung melody in my DAW, where I feel more confortable working. Then I import it into Vocaloid, typing the lyrics. This is the first step, a rude draft of the song. Press play and shit happens everywhere.

Next I do is working in pronunciation flow, making syllabes to sound more natural. Usually English vocaloid banks tends to sound too much syllabic.

Mostly you have to slice a syllabe into 2 notes to make the vowel longer. Sometimes you need to do the contrary, making the voiced consonant before the vowel longer. ( "I caaaaaaaan't" or "I cannnnnnnn't" ). I always compare the waveforms of the vocaloid render vs the reference track in the DAW to fix those problems. There is a lot of tricks to force the editor to do what you need or fix what you don't, like DB errors or missing diphone combinations. Old dog here.

Next step I do is to work in expression performance. Scoops are the soul of singing, without them it would sound like a piano or an untrained singer. Using timestreching over the reference track (original or your own recorded voice) in slow down tempo would be very helpful to notice the scoops, vibratos and dynamic changes more accurately to recrerate the singing performance. If you still have problems, compare the pitch of both tracks to see what's going on.

Vibrato editing is a nightmare due their abstract representation. To edit and listen over and over is a must until you get what are you are looking for. Each attempt takes some seconds plus listening, make your numbers.

Finally I fix dynamics. Depending of the song, not much is needed, just to compensate energy in soft or strong syllabes, by ear. Obviously you can't do miracles, is imposible to transform a soft voice in a strong and tense one or viceversa,but now if the voice bank has different databases (whisper, strong, etc) is posible to mix them via cross synthesis (I don't tried it yet).

As as said, Vocaloid is not mature enought to take it seriously in music production until you are crazy, passionate, or you have a lot of spare time to edit. Good for some chorus lines hidden in the back.

But there is a lot of good researchers working in singing synthesis today, so you can expect good news and huge advances in any moment. I do.

Kudos for the cogent description of a working process. I see too little of this.

[SODDI]

dupe, sorry

[Xenobioz]

Does anybody use Utau? It's like vocaloid but free. It's quite difficult to use and get working since it's a Japanese program.

I've heard of it. I think it's in English too. Have you had any success with it?

Yes, I have the English version. A few things are still in Japanese though. The hardest thing is entering lyrics. Most of the Utaloids use Hiragana or Katakana. You can also make your own Voicebanks which can be fun.

If it could be run as a VSTI I would probably use it more. Now I haven't used it at all since I tried it the first time.

[provoc]

Hi All,

I used UTAU for the chorus vocals on a track I just posted so it is there to have a listen to as an example.

http://www.kvraudio.com/forum/viewtopic ... 4&t=433348

It worked for what I wanted since I wasn’t after realism in fact I considered vocoding the vocals to make them more synthetic but intelligibility was already an issue so I didn’t want to push it since they already might be hard to understand, especially in a crap listening environment.

As Xenobioz mentioned because it is not a VST it is a bit of a pain, especially to get the timing right and in my case I just got close enough within UTAU and then rendered for reaper and cut it up and timestretched there it to get it to fit.

A couple of other observations. I used the voice bank Mimi Yorune and some of the phonemes have not been set up correctly so they are not edited to the correct start and finish so you have to either fix them (tedious) or choose another vowel combination that is close enough.

Definitely a pain to edit but I believe Vocaloid is too. Not really too bad for what I did since I didn’t have a lot of lyrics but doing a whole song would require a lot of practice and patience!

Also the English wiki and docs are pretty good and the interface itself is piano roll based with reasonable editing options and a pretty decent feature set including vibrato, pitch bending etc. But working out the phonemes is fiddly, but if you look up CZloid on the internet she has some good tutorials.

Bottom line for me as a hobbyist is that it got the job done especially since I like to use free software as much as possible to see what I can achieve with it For a pro who wants things done quickly then his/her experience would likely vary.

Cheers,

[PaulB23]

Just got vocaloid 4 cyber diva a week ago and quite impressed with what it can do out of the box - picked up avanna too as it was on sale.... Nothing to match the fantastic work posted above - but my first go at using Vocaloid on Soundcloud is here:

https://soundcloud.com/orgonetic/alchemy

This was cyber diva + avanna doubled up.

The editor is not very intuitive to a newbie - Difficult to get the phoneme / articulations as clear but quite like the synthetic sound anyway... Will be working on how to get it more understandable

Not yet worked out a simple way to get vowels to span different notes - I guess you have to break the phonemes apart yourself rather than using the insert lyrics feature that automatically decomposes on the syllables ?

[stikygum]

Just got vocaloid 4 cyber diva a week ago and quite impressed with what it can do out of the box - picked up avanna too as it was on sale.... Nothing to match the fantastic work posted above - but my first go at using Vocaloid on Soundcloud is here:

This was cyber diva + avanna doubled up.

The editor is not very intuitive to a newbie - Difficult to get the phoneme / articulations as clear but quite like the synthetic sound anyway... Will be working on how to get it more understandable

Not yet worked out a simple way to get vowels to span different notes - I guess you have to break the phonemes apart yourself rather than using the insert lyrics feature that automatically decomposes on the syllables ?

This is probably one of the best sung/pronounced songs I've heard with Vocaloid. Nice work! I didn't know Avanna had a version 4 out.

[wagtunes]

Just got vocaloid 4 cyber diva a week ago and quite impressed with what it can do out of the box - picked up avanna too as it was on sale.... Nothing to match the fantastic work posted above - but my first go at using Vocaloid on Soundcloud is here:

This was cyber diva + avanna doubled up.

The editor is not very intuitive to a newbie - Difficult to get the phoneme / articulations as clear but quite like the synthetic sound anyway... Will be working on how to get it more understandable

Not yet worked out a simple way to get vowels to span different notes - I guess you have to break the phonemes apart yourself rather than using the insert lyrics feature that automatically decomposes on the syllables ?

This is probably one of the best sung/pronounced songs I've heard with Vocaloid. Nice work! I didn't know Avanna had a version 4 out.

I'm still waiting for Diva and the Cubase editor to be released in the US. If it's not by the end of the year (the site says 2015) I'm just going to take my chances with Ebay as much as I dread doing it.

[stikygum]

Is there a list somewhere of all the Vocaloid 4 singers? I want to do some research before I choose one.

[arkmabat]

Just got vocaloid 4 cyber diva a week ago and quite impressed with what it can do out of the box - picked up avanna too as it was on sale.... Nothing to match the fantastic work posted above - but my first go at using Vocaloid on Soundcloud is here:

https://soundcloud.com/orgonetic/alchemy

This was cyber diva + avanna doubled up.

The editor is not very intuitive to a newbie - Difficult to get the phoneme / articulations as clear but quite like the synthetic sound anyway... Will be working on how to get it more understandable

Not yet worked out a simple way to get vowels to span different notes - I guess you have to break the phonemes apart yourself rather than using the insert lyrics feature that automatically decomposes on the syllables ?

Thanks for sharing. Very interesting.

[arkmabat]

It seems like it's a great songwriting tool. And if you really had something good going, you could pull in a real singer later.

[andrew732]

It seems like it's a great songwriting tool. And if you really had something good going, you could pull in a real singer later.

This type of mock-up is what I use it for, although I find that even with the ability to (sort of) pronounce English words, it's not much better in that capacity than the generic "ahhhs" and "lahhhs" that can be found in, for example, Omnisphere. Also, the English language support and even the ability to actually purchase it from a website in English has been minimal or non-existent in the past. This was a few years ago, but I remember it took me weeks with Google Translate to actually track down the only English vocaloid available at the time, only to discover that there was virtually no documentation in English. Maybe that's changed now, but if the OP is interested in expanding outside of Japan, they could start by addressing that.

[wagtunes]

It seems like it's a great songwriting tool. And if you really had something good going, you could pull in a real singer later.

This type of mock-up is what I use it for, although I find that even with the ability to (sort of) pronounce English words, it's not much better in that capacity than the generic "ahhhs" and "lahhhs" that can be found in, for example, Omnisphere. Also, the English language support and even the ability to actually purchase it from a website in English has been minimal or non-existent in the past. This was a few years ago, but I remember it took me weeks with Google Translate to actually track down the only English vocaloid available at the time, only to discover that there was virtually no documentation in English. Maybe that's changed now, but if the OP is interested in expanding outside of Japan, they could start by addressing that.

Vocaloid has come a long way since then. The quality is no longer the issue. The issue is purchasing in the states. Hopefully, that will be resolved soon.

[stikygum]

Ok, I think I'm going to buy Vocaloid Editor for Cubase and CyberDiva today off of ebay. I tried listening to some version 4 Vocaloids, but think CyberDiva might have the clearest sound, I'm not sure though.

[wagtunes]

Ok, I think I'm going to buy Vocaloid Editor for Cubase and CyberDiva today off of ebay. I tried listening to some version 4 Vocaloids, but think CyberDiva might have the clearest sound, I'm not sure though.

It is not ITB clear. I've heard some horrible Cyber Diva demos. There is no question that it is going to take a fair amount of skill to get this thing to sound as good as some of the demos. The disparity in the quality of the demos out there is living proof of this.

Expect to spend quite some time on this getting it to sound "good".