Forum

Please consider registering
guest

Log In

Lost password?
Advanced Search

— Forum Scope —




— Match —





— Forum Options —





Minimum search word length is 4 characters - maximum search word length is 84 characters

The forums are currently locked and only available for read only access
This topic is locked No permission to create posts
JMT PM algorithms can now control Vocaloid “Vibrato Objects”
May 18 2004
11:12
alexeiustinov
Member
Members
Forum Posts: 34
Member Since:
Apr 8 2004
Offline

Hello Vocaloid community!

Thanks to everybody for your interest and support. Our new step: Vocaloid’s “Vibrato Objects” are automatically added by special PMStyle – Look at the Vocaloid *.mid files. Also, YV Enhancer 1.0 now is under development, and we hope to hear your advice during its finalizing. Together we will force this thing to sing! :D As always, no manual editing at all! See details at [url:1d3zsys5]http://www.ntonyx.com/vocaloid/05-18-2004.htm[/url:1d3zsys5]

Alexei Ustinov,
CTO, JMT

May 18 2004
13:59
robotarchie
Member
Members
Forum Posts: 223
Member Since:
Feb 27 2004
Offline

Having played the mp3 a few times I still just don’t get it. There are parts of the demo that are interestingly done to be sure, but the first thing I noticed was that the voice fades in and out so wildly it reminded me of a drunk japanese businessman on stag karaoke night with a very bad mic technique. I’ve heard other people praise your software and wonder whether I’m missing something – or is this the Emperor’s New Code? Sorry to be so harsh, but this just doesn’t float my boat at all. I can’t see the benefit of it. Perhaps the voice (as an instrument) just doesn’t suit being heavily stylised in the way that a saxophone or whatever other instrument might? Sorry….

May 18 2004
22:24
bobyordan
Member
Members
Forum Posts: 88
Member Since:
Feb 28 2004
Offline

I think this new demo shows great progress, but there is still some fine tuning to make Leon sound more natural. Considering this is autogenerated, it is fantastic so far. :)

The many small shades that a voice contains is not the easiest of tasks to capture correctly at once. Then there is the phonetic raw material that perhaps also needs a tweak?

The voices sound much more open now, than before.

I rather have a drunk japanese buissness man/woman singing my tunes within a couple of minutes, than tweaking for 8 hours and still not feeling 100% satisfied. <img class=” />

One thing that I hear in this demo is that Leon sounds okay on long notes,
but on most of the short phrases make a staccato impression. There is to little “flow” in those phrases. The words should glide into each other more than they do now.

Sometimes Leon makes a small “gurgling” sound, could have to do with the phonetic phrases? Sounds a bit like singing with the mouth partly beneath water. ;)

I have noticed that some words like eg see & dream should need open mouth high frequency sound during the e. Now the frequency boost comes at the end of the word together with the vibrato. Perhaps if the frequency boost starts generally in the middle of the word?

Cool song by the way, I like it. :)

May 19 2004
03:54
alexeiustinov
Member
Members
Forum Posts: 34
Member Since:
Apr 8 2004
Offline

Dear Archie !

If I have correctly understood your comment I should tell the following (and probably, it will be interesting to other users too):
1) «Type your words along your tune and you get excellent vocal track» is a bluff. It’s impossible.
2) Traditional (acoustic) scheme of musical communication has a very big difference from electronic scheme. Please look:
“paper score” => “performer (a person)” => “instrument” => “real sound”
and
“electronic score (*.mid)” => “sequencer” => “synth” => “real sound”
The first scheme has a performer that produces not only score, but “expressive elements”,
and the second one hasn’t got such a component.
3) Can a program replace a live musician? Yes and No!
Why “yes”? Modern synths/samplers can produce ANY sound. However, even Stradivarius violin or Caruso’s voice short wave sample (or spectrum) will morph into “car horn” during playback on the loop. So we have to manage it accordantly to psycho-acoustic laws. Also: differential thresholds of human reception are easily defined, and existing synth devices can overcome them without any special problems. In other words, it is possible to create an example when a pro musician won’t be able to recognize its artificial nature.
Why No? Because for now we don’t know how to manage synth in a right way.
4) So, we can never get anything sounding like real voice without adding “expressive elements” into electronic score (*.mid). How do you offer to do it? Even musicians (a singer, a guitarist, …) don’t know what they do! Do you want to ask them about ms, dB, cents, pitch, breath curves, etc.? How you will add all this into your Vocaloid *.mid? OK, you can use real soundtracks of world-class singers, apply a lot of measuring software (pitch, volume, spectrum detection) and spend a huge amount of your efforts to find out their Pitch, Breath curves and spectrum samples to bring all this into your tune. Go for it!

About your doubts. If you have the right vision of “an order of things” you are welcome to make any attempts to achieve result. BTW, the first airplane was by many considered useless, and the director of first big IBM computers company said that the demand for them won’t be more than 5-6 machines a year, etc.
About our work. I didn’t say that we know how it will be done, but I said we know where we are going.

May 19 2004
04:38
alexeiustinov
Member
Members
Forum Posts: 34
Member Since:
Apr 8 2004
Offline

Dear Bob, dear Archie

1) “Drunk” singing is a result of big time Pitch smoothing that can be easy changed.
2) “Staccato manner” is a feature of YV synth, however it also can be adjusted.
3) “Breathing” or “With a Pressure” singing depends on the values of Resonances parameters, but it is necessary to manage them simultaneously. It is our next task.

Actually we are working on the “parametrical morphing”, and only when YV Enhancer is done as an alpha version, I will have some freedom for the effective testing of YV synth. All our examples only show that “expressive elements” can be calculated by a program. Of course, the developing of Performance Modeling algorithms is a separate and quite hard task. Our algorithms are not excellent, but I don’t know anyone who can show anything better. However, with Vocaloid we have quite new situation. We will develop PM algorithms “for the exact certain executive device”, so every user of Leon / Lola will have the same result. Unfortunately, when we developed existing PM algorithms (SE30/40, Sonar Plug-in, Onyx) we didn’t know by what synth this controllers will be managed. What you think, is it possible to train a pianist without instrument? :D

Have success in all you do :D

May 19 2004
07:32
bobyordan
Member
Members
Forum Posts: 88
Member Since:
Feb 28 2004
Offline

Hi Alexei

Thanx for you answers. But I disagree with one of your statements
“Our algorithms are not excellent”. Because they are, probably not ultimate, but excellent. I use your algoritms in almost every song I make, to bring expressive enhancements. :)

They can breath life into a otherwise dull melody line or sound etc etc.

I got one question to you. Are the lyrics saved within the midi file? If so is it possible to include an “import lyrics from text file” feature into the YV enhancer? Like if I include separators in a text file to mark what phrase should be placed on a note.

eg

You are the sun#shine of my life

Are placed on the first 8 notes in the melody line.

And if I specify that the import should be to starting from the first note in bar 50 of the song.

%50 Yes#ter#day all my troub#els seem so far a#way

or

%10&50&70 Yes#ter#day all my troub#els seem so far a#way

The same phrases are placed on three places in the song.

Otherwise it would be cool to have that feature included in the vocaloid.

May 19 2004
08:11
robotarchie
Member
Members
Forum Posts: 223
Member Since:
Feb 27 2004
Offline

Hi, Alexei. Carry on, go for it, good luck. Yes, it’s a very complex problem, and the cross-babble of “language” can often generate confusions of reference and meaning when marketing men, musicians, academics and/or scientists meet to discuss the practical aesthetics of sound. It’s a rare animal indeed that is all of those things. When you say “parametrical morphing” are you describing a “roving tone control”? My criticism of your company’s demos to date is simply that to my ears (the pink rubbery things on the sides of my head) they produce an overall effect that is a tad “over-cooked”. Play the demo to someone who is listening without prejudice and see eyebrows being raised…

Your approach (as it appears to me to date) is one of creating a new and better interface between the blank editor page and the raw tools of control found in vocaloid’s MIDI implementation chart. As we know, working with vocaloid via hand coding is like preparing a rather complex gourmet meal. It takes a lot of knowledge, experience, skill, patience and time to come up with a half decent new dish. Maybe it takes too much of all these things when the results are so seldom truly satisfying? Your software appears to address this perceived problem by adopting a microwave ready-meal approach. It ambitiously strives to replace the tedium of the chef’s knife and chopping board with an electric blender. A “Magic Mix”?

Below is a quote from user Gray from another thread on this forum entitled “Are You Happy with Vocaloid?”. He says, quote:

“I found that most of my problems were due to overuse of parameters. Had to try them all out though to see what they did, since the manual didn’t give much of a clue. Lemme see, Harmonics?? Thats a fancy name for volume control. heh heh. Velocity?? works on volume somewhat, but if you wish to fade out better use harmonics instead. Noise- only needed when you wish to accent the start of a word for better pronunciation. Gender change?? used to find a spot that sounds right in the key your in. If you really wish to change gender, best do an octve higher or lower…”

May 19 2004
08:12
andromeda
Member
Members
Forum Posts: 188
Member Since:
Feb 27 2004
Offline

An interesting discussion here. I note, Alexei, that you say you are going to try to find a way of adjusting the resonances in the vocaloid voice, using your software. As I have said on here many times, this, to me, is what has been needed all along. The sooner the better. I agree that getting performance elements into a vocaloid sequence is going to be difficult. The scale of the problem is huge. The subtlety of the human voice and its possible inflections is an order of magnitude greater than any other musical instrument. The sensitivity of the human ear to these inflections is also much greater. Most people listening to a vocal track will of course not be consciously analysing the parameters. They will just feel that the voice sounds “right” (a vocaloid voice). However, for a musician to program the vocaloid voice to “sound right”, will he/she have to understand the subtleties? The way the vowel sound changes, and the resonances change. Is it possible to program a vocaloid so that it can do all this and “sound right”. Yes, either by the musician doing a careful study of vocal technique, combined with sound analysis graphs etc, or your software providing “templates” which give these ready made. As soon as you start to use templates, you limit the expression to certain pre-defined examples, and consequently limit your scope for expressing exactly what you wanted to express. What I’m saying is, the path you are on may eventually lead to something useful, but my guess is you are a couple of years, at best, away from doing what really needs to be done. It’s not the expressive elements I need. I need to voice to “sound right”. Good luck with your project. See you again in 2006.
Chris

May 19 2004
08:23
robotarchie
Member
Members
Forum Posts: 223
Member Since:
Feb 27 2004
Offline

Andromeda – I think that when the imminent (?) vocaloid update/fix allows the VSTi to work proerly, you’ll be able to tweak the resonance parameters from within your familiar VST compliant host. This will (probably) be in realtime and be capable of being put under mix automation control. It’s still going to be a bit D.I.Y, but I’m used to having control over my other synths, so I don’t think I’ll be switching my host sequencer allegiances right now for the sake of an easier (vocaloid) life….

May 19 2004
10:17
andromeda
Member
Members
Forum Posts: 188
Member Since:
Feb 27 2004
Offline

Archie, I hope the new Vocaloid version does allow real time tweaking of these parameters.. Personally, I would prefer this within the Vocaloid editor itself, rather than via VSTi. i.e. using the synthesize while play, control (or whatever it’s called). I think I was a bit optimistic when I said it would take 2 years to get the voice sounding “right”. Vocal synthesis at the moment seems to be at the same point that instrument sampling/synthesis was in the 80s, in my view. I’m not saying it will take 20 years though. My dream would be a Vocaloid keyboard where you had control of all the parameters via knobs etc and just typed in the text. Playing the keyboard would “play” the song- including harmonies and different text for different parts. Layering could produce a choral effect. Yes I know this could all be set up on a PC, but I would miss the real keyboard. I really don’t like the claustrophobic feel of working on a PC. (We’ve discussed this before on here, I know). I think the raw phoneme (sampled) input is also very important. The sound you produce must inevitably be heavily dependant on the original voice used. I find Lola sounds much better and more natural when used in the low range where she was sampled and where, I would guess, her voice was most expressive.
Cheers
Chris

May 19 2004
21:44
gray
Member
Members
Forum Posts: 304
Member Since:
Feb 27 2004
Offline

Bob, I install lyrics on top of my melody before loading in vocaloid. Must have no spaces / , . etc between words. Just words in lyric format. If the words are placed properly with the notes and there are no note overlap, the lyrics will load for you. Then all you have to do is the phoneme translate in vocaloid. The bad side is all my tracks I did lyrics on were in text format. So I have to change them all to lyric for vocaloid. Any notes overlapped, the lyrics will be in the wrong place.

May 20 2004
08:54
bobyordan
Member
Members
Forum Posts: 88
Member Since:
Feb 28 2004
Offline

[quote="gray":3hdc8nz6]Bob, I install lyrics on top of my melody before loading in vocaloid. Must have no spaces / , . etc between words. Just words in lyric format. If the words are placed properly with the notes and there are no note overlap, the lyrics will load for you. Then all you have to do is the phoneme translate in vocaloid. The bad side is all my tracks I did lyrics on were in text format. So I have to change them all to lyric for vocaloid. Any notes overlapped, the lyrics will be in the wrong place.[/quote:3hdc8nz6]

Hi Gray

Sounds great! But I dont understand how you enter the lyrics?? Do I need a special midi editor that has a lyrics format input mode? :)

Wonder if it is possible to remove overlaps automatically in any of my midi
programs (NTONYX, Cakewalk MC, BIAB 2004,Cubase SX, Pro Tools FREE) ? Must check the manuals. :)

May 20 2004
10:09
administrator
UK
Admin
Forum Posts: 268
Member Since:
Feb 4 2004
Offline

[quote="BobYordan":z04q8s6q]Wonder if it is possible to remove overlaps automatically[/quote:z04q8s6q]

on my version of cubase (VST32 5.1) it is called “restrict polyphony” (under Midi functions) and you just set it to 1.

May 20 2004
13:03
robotarchie
Member
Members
Forum Posts: 223
Member Since:
Feb 27 2004
Offline

Gray – I’m confused again (that phrase will be on my gravestone, no doubt). Like Dromie I’d like a clearer explanation of what you mean when you say you load the lyrics before (?) vocaloid. I didn’t know you could do that? How does that work? I thought all lyrics had to be typed directly into the vocaloid editor… ?

May 20 2004
13:21
administrator
UK
Admin
Forum Posts: 268
Member Since:
Feb 4 2004
Offline

vocaloid will open lyrics that are embedded in a midi file…. is that what you mean?

May 20 2004
13:37
robotarchie
Member
Members
Forum Posts: 223
Member Since:
Feb 27 2004
Offline

Yes – that’s a new one on me! I’ve been using Cubase SX for yonks and never knew such a thing existed – but then I never had a vocaloid till recently. I must have a wee poke around in SX pretty soon and see how that can be made to work. It would speed things up (probably). Maybe this explains a little of how Gray can pump out tunes like a mad thing on mad pills. Interesting development ….. 8O

May 20 2004
21:23
bobyordan
Member
Members
Forum Posts: 88
Member Since:
Feb 28 2004
Offline

[quote="Administrator":2jsyqlap][quote="BobYordan":2jsyqlap]Wonder if it is possible to remove overlaps automatically[/quote:2jsyqlap]

on my version of cubase (VST32 5.1) it is called “restrict polyphony” (under Midi functions) and you just set it to 1.[/quote:2jsyqlap]

Thanx admin, I will check it out in SX 1.0.6 Hopefully there is a simular feature there? :)

Cheers
Bob
:roll:

May 20 2004
22:12
robotarchie
Member
Members
Forum Posts: 223
Member Since:
Feb 27 2004
Offline

Yes again, Joe. Fuck me, I just want to know the details…. <img class=” />

May 20 2004
23:42
gray
Member
Members
Forum Posts: 304
Member Since:
Feb 27 2004
Offline

I use midilacator with the 15 buck upgrade for piano roll. the sequencer is free at midilocator.com. It has options to enhter lyrics in either text or lyric format. There are scads of karaoke files out there with lyrics already entered. Just copy the lyric track to the melody track and it’s vocaloid ready after breaking the track into tiny little bits of course.

May 21 2004
09:03
alexeiustinov
Member
Members
Forum Posts: 34
Member Since:
Apr 8 2004
Offline

Hello!
Sorry for my absence but as always, I am very busy :) [quote:2aao7722]Bob Yordan wrote:
Are the lyrics saved within the midi file?
If so is it possible to include an “import lyrics from text file” feature into the YV enhancer?[/quote:2aao7722]
This is a very simple task. Though often it’s the simple things that you don’t have time for. In the nearest future we will post a poll on our web site regarding the YV Enhancer functionality, to take into account our users’ demands.

[quote:2aao7722]Robot Archie wrote:
When you say “parametrical morphing” are you describing a “roving tone control”? My criticism of your company’s demos to date is simply that to my ears they produce an overall effect that is a tad “over-cooked”.[/quote:2aao7722]
Yes, we reinforce the impact deliberately for EVERYONE to notice. Bit the algorithms themselves, if you are familiar with Modeler (SE, Onyx) allow to reduce ANY parameter to zero. Surely, the impacts should be finer. And these fine impacts, as condiments, are not always noticed when they are there, but without them everything becomes plain and tasteless.

[quote:2aao7722]Andromeda wrote:
As soon as you start to use templates, you limit the expression to certain pre-defined examples, and consequently limit your scope for expressing exactly what you wanted to express.[/quote:2aao7722]
Don’t agree.
We can’t entrust everything to the program yet. We have to control the result, as an orchestra conductor or a film producer. You don’t want to use automatically a filter in PhotoShop or CorelDraw when editing, say “Red eyes effect” – otherwise cherries and apples in your photo will become gray! The primary PM task is to basically enliven the vocal, and only after that can we talk about any specific performer’s manner. So you shouldn’t worry about the template that limits your creative freedom.

[quote:2aao7722]Robot Archie dreamt:
My dream would be a Vocaloid keyboard where you had control of all the parameters via knobs etc and just typed in the text. [/quote:2aao7722]
1) It seems to me that calculating the vocal wave is way too demanding and laborious to speak of real time control.
2) Real simplification of editing (either in Vocaloid itself, VST-plug-in or with YV Enhancer) will be in the possibility to adjust all “expressiveness elements” without lyrics (i.e. without phoneme syntheis)in some simple wave, say, “ah-ah-ah”, but with all the controllers engaged. And only then, when all the elements are adjusted, add the lyrics. BTW, this is how professional vocalists learn the songs!

This topic is locked No permission to create posts
Forum Timezone: UTC 0

Most Users Ever Online: 108

Currently Online:
5 Guest(s)

Currently Browsing this Page:
1 Guest(s)

Top Posters:

gray: 304

robotarchie: 223

andromeda: 188

quetzalcoatl: 173

Giuseppe: 167

Luka Mitutoyo: 147

Member Stats:

Guest Posters: 1

Members: 7988

Moderators: 0

Admins: 1

Forum Stats:

Groups: 5

Forums: 27

Topics: 764

Posts: 3555

Newest Members: kprem, yang7764, lisajim, gameace, Sunnydoot, Anildoot

Administrators: administrator: 268