V-Vocal coloring vocals

When I use V-vocal on a vocal clip, it colors the sound, even on unaltered notes.  In other words, if I change one syllable, the entire clip has a very slight, but noticeable "robotic" or "hollow" sound.  It is especially apparent in breaths and consonants such as F's and S's.
 
Does anybody else notice this?  Is there a setting I may have missed to counter this?
 
(Please, no wiseacres with the "use melodyne" retorts--although I'm seriously considering getting it)
Cheers all!

Are you highlighting the vocal part in v-vocal you are editing? and are you 1000% sure your not editing the entire clip

use melodyne...
in all seriousness, though DO use melodyne.

Any tuning programme is going to colour the sound, its using an algorhythm to process the effect. what bit rate did you record and what sample rate? really if you're doing pitch or time shifting go for 24bit and at least 48k probably higher if you're system will handle it, gives you more headroom for that kind of processing.
Also, if you're finding this problem when you only want to alter things slightly, just split the portion you want to use and V-Vocal that part!

Hi CJ!  I'm not sure what you mean about highlighting in V-Vocal.  I just move a note and it highlights that vertical section automatically.
 
I suppose I cannot be 1000% sure about anything.  Is there some implicit editing of an entire clip when you just move a few notes?
 

I record at 24bit/48k...

cho_drummer


use melodyne...
in all seriousness, though DO use melodyne.

Any tuning programme is going to colour the sound, its using an algorhythm to process the effect. what bit rate did you record and what sample rate? really if you're doing pitch or time shifting go for 24bit and at least 48k probably higher if you're system will handle it, gives you more headroom for that kind of processing.
Also, if you're finding this problem when you only want to alter things slightly, just split the portion you want to use and V-Vocal that part!

melodyne (well the plug in version at least) still has an effect on the tone of the vocal, albeit slight...

just a point, back to our regularly scheduled program...8)

cheers

Wiz

Quick list of things that can cause the sound you are describing in fact many people try to get that robotic effect.

Try reducing the pitch follow and sensitivity controls.

On the formant screen you can adjust those sections up and down. Select the area and then move the line. The auto process will get you close but you may need to tweak "problem" areas. This is where you can do it.

If you want to change th entire clip you could raise or lower the formant line without selecting part of the clip.

Hope that helps.

Much appreciated advice Robin, and it will definitely come in handy in when editing vocals.

But I am talking about unedited portions of a V-Vocal clip.  For example:  right-click a clip, select "Create V-Vocal Clip" and stop right there.  The clip (in my experience) has already been polluted at this point.  It's very subtle coloring, but my ears can definitely hear something unnatural and a bit hollow (possible phasing??).  Especially in breaths, F's and S sounds.

That Color you hear could be a phasing if there is background noise on the track...It could also the the on-line processing algorithm. If you create a vvocal clip and make no changes to it and then bounce it to a clip does it sound altered in anyway?

itllcometogether


Much appreciated advice Robin, and it will definitely come in handy in when editing vocals.

But I am talking about unedited portions of a V-Vocal clip.  For example:  right-click a clip, select "Create V-Vocal Clip" and stop right there.  The clip (in my experience) has already been polluted at this point.  It's very subtle coloring, but my ears can definitely hear something unnatural and a bit hollow (possible phasing??).  Especially in breaths, F's and S sounds.

Yes yes - we've reported this many times along with the periodic crashing problems associated with VV. I wish Cakewalk would chime in here already. Maybe they don't because it's not their product. Call Roland
 
The argument about background noise has merit but I don't have that problem with Auto-tune. If AT can deal with a little background noise (actually, it's amazing how forgiving AT is of background noise) then VV should be able to handle it too. Actually, it used to. This nonsense crept in along the way and hasn't been fixed yet. We shouldn't have to record in an anachoic chamber just so VV can track properly. And yes, I know, it imparts some unpleasantness even on portions of the audio that aren't being fixed.
 
VV has it's strengths though and I wish it would work as expected because if it did I'd use it more often. I used it a lot when it first was introduced. The thing I don't particularly like about AT is that you have to play the section of music you want to process first (it encodes the music during playback). This is all real-time work. With VV, the encoding happens in seconds. You can encode a one minute audio clip in a few short seconds and begin work. That makes a big difference when you have many tracks of audio to deal with. I rarely, if ever, use AT in automatic mode. It almost never produces 100% perfect results that way. AT's user interface is a little clunky compared to VV. I like the way you can scale VV to the screen size. The workspace is a lot easier to see and navigate around in compared to AT.
 
I'd consider Melodyne if I didn't already have AT. It's not like I make a living pitch-correcting vocal tracks. It's just a tool for occasional work. I guess if I was doing hip hop or urban R&B, it would be another story. Every rap or R&B song released in the past two years has AT all over it. Everyone sounds like a robot. I guess robots really have taken over planet earth

itllcometogether


When I use V-vocal on a vocal clip, it colors the sound, even on unaltered notes.  In other words, if I change one syllable, the entire clip has a very slight, but noticeable "robotic" or "hollow" sound.  It is especially apparent in breaths and consonants such as F's and S's.
 
Does anybody else notice this?  Is there a setting I may have missed to counter this?
 
(Please, no wiseacres with the "use melodyne" retorts--although I'm seriously considering getting it)
Cheers all!

V-Vocal has the *potential* to color the sound as a result of its encoding process (the VariPhrase technology originally debuted in the Roland VP-9000 hardware sampler).  Certain types of signal coloration introduced during tracking can cause V-Vocal's own coloration to be more pronounced; in general, the more pristine the source material, the more true-to-original-color V-Vocal will be.  You can find many, many, many threads about this here on the forum if you search; but you'll have to make sure to set the forum to "All Topics" (instead of just the last 30 days) before searching, as some of the threads are quite old at this point.

Cakewalk did chime in (me). At first I thought it was something in the process of using the plugin hence my post. Since then i have reported back the results to the product team.

Unfortunately, I can't give you a play by play on the next steps as I am not involved in them, I am a sales guy (duck). I can tell you that the right people are aware of this thread and what you have reported.

Robin

Robin Kelly [Cakewalk
]

Cakewalk did chime in (me). At first I thought it was something in the process of using the plugin hence my post. Since then i have reported back the results to the product team.

Unfortunately, I can't give you a play by play on the next steps as I am not involved in them, I am a sales guy (duck). I can tell you that the right people are aware of this thread and what you have reported.

Robin

Thanks for the feedback, Robin.  As noted, the typical V-Vocal coloration manifests as a "phasey" or "flangey" sound, considerably more prominent on sibilance than on other portions of the vocal.  It can also make the vocal in general sound "hollow."  It seems to be extremely dependent on how pristine the source material is - the less coloration in the source (i.e., higher-quality microphone, higher-quality preamps, higher-quality converters), the less problem with V-Vocal adding more coloration.  The more coloration in the source (i.e. the average dynamic microphone, noisy preamps, lower-quality converters), the more chance of V-Vocal adding more coloration.
 
Melodyne, as well as newer versions of AutoTune, seem to have come up with ways of approaching their encoding of the audio that minimize these types of coloration.  I don't know whether Roland has been working on updating their encoding approach, but it seems like the type of thing that is deeply "under-the-hood" on Roland's end, as opposed to a simple fix from Cakewalk.

But I am talking about unedited portions of a V-Vocal clip.

Caveat: I have almost no experience with V-Vocal, but... if you know in advance that you only want to fix that one syllable, maybe you could split it out of the rest of the clip before applying V-Vocal...? Just a thought for this particular situation.

itllcometogether


Much appreciated advice Robin, and it will definitely come in handy in when editing vocals.

But I am talking about unedited portions of a V-Vocal clip.  For example:  right-click a clip, select "Create V-Vocal Clip" and stop right there.  The clip (in my experience) has already been polluted at this point.  It's very subtle coloring, but my ears can definitely hear something unnatural and a bit hollow (possible phasing??).  Especially in breaths, F's and S sounds.

When I use V-vocal I split clip so that only the small sections I want to edit loads into the application

V-vocal keeps an unedited copy of the waveform muted and beneath the new V-vocal clip so if you've loaded an entire clip into it and then realise you only need part of it, you can split the the Vvocal track, delete the part you don't need and then unmute the original recording (highlight it and press Q), then just make sure the parts you have edited in this clip have either been slip edited or highlighted with the mute cursor (k on the keyboard and then highlight the parts in the take you want to mute)

hope this helps you in someway, sorry if it doesn't
all the best

Jon

Tom Riggs

That Color you hear could be a phasing if there is background noise on the track...It could also the the on-line processing algorithm. If you create a vvocal clip and make no changes to it and then bounce it to a clip does it sound altered in anyway?

Thank you all for the on-target feedback!  Tom's question was very helpful and I've been experimenting.  It is looking like an issue with V-Vocal's realtime processing ability.  When I bounce a V-Vocaled clip and A/B, the phasey stuff is much less apparent.
 
Perhaps this is related to other Sonar settings, like latency, or dithering?  Any other ideas are welcome.  Thanks again!

V-Vocal should have no effect on unedited portions of the clip. If it does, that's a bug.

Try this experiment:

1. Create a V-Vocal clip. Do not make any edits, just close V-V.
2. Bounce the V-V clip. You now have two clips that should be identical, assuming V-V made no alterations.
3. Move the bounced clip to another track, using SHIFT-drag so it remains precisely in the same time location as the original.
4. Un-mute the original clip by pressing "Q".
5. Invert the phase of one of the clips and then bounce the two to a new track.

If V-Vocal modified the original clip in any way, the difference will be in the final bounce. If no modifications were made, the bounced clip should be a flat line.

I've actually done this test, but I don't want to give away the surprise. Try it.

Great experiment bitflipper!  Can I spoil the surprise?  Assuming you had the same surprise as me:  the untouched, bounced V-Vocal clip is far from identical to the original clip!

This is a seriously dirty plug in.  With all the time, money, and equipment we all put into getting the best possible vocal recording, this pollution is not acceptable. 

I do love the tool, and I sincerely hope Cakewalk and Roland take this seriously.

You're absolutely right. The picture below shows an extreme (sample-level) zoom view of the original on top, and the bounced data from V-Vocal on the bottom:



Clearly, they are not identical. For starters, every sample has been shifted by one sample - but that can be forgiven, since as long as the two aren't combined it shouldn't have any noticeable effect. But the spectral content has also been altered a great deal, with the corruption most severe in high frequencies. This particular portion is a closeup of an "S" in a vocal clip.

By contrast, the next image is from another part of the same clip that did not have much high-frequency content. Note that aside from the 1-sample offset, the two are nearly identical. (Ignore the arrow, it's just where the cursor was when I took the screenshot)


But less than a second later, the corruption is clearly visible:



NOTE: In the above screenshot I have nudged the bounced copy back 1 sample so the two would be in phase and therefore easier to visually compare.

Glad to see others noticing change of sound with Virtual Vocal. A quote from this site says
Perfect vocals – on pitch and in time.

I get anything but perfect vocals. Even tiny corrections of pitch make the vocals sound horrible. They get this synthetic colouration that is very noticeable and renders it unusable for serious work.
I'm not talking about a half tone correction, I would never even try that.

Not to mention the audio to midi function, that produces a result that has absolutely nothing to do with the simple file I try to convert.

This area of sonar is way behind the ballpark , and needs a serious overhall in terms of user interface and achievable results

Actually, I use V-Vocal all the time and get perfectly acceptable results. But you do have to use it on its terms. It has lots of room for improvement, but it is usable.

I really like V-Vocal and I've been using it for a few years with great results.  But it has rendered many vocal clips unusable with phasey artifacts.

Don't get me wrong; the phasing is very slight, not as if it sounds like Mr Roboto.  But I did some blind listening tests and I could tell 100% of the time when V-Vocal was active vs bypassed on certain clips.  And this was even with zero correction applied.

Im still waiting for v vocal to reach melodynes level. melodynes sounds better, period-

sotbee


I get anything but perfect vocals. Even tiny corrections of pitch make the vocals sound horrible. They get this synthetic colouration that is very noticeable and renders it unusable for serious work.

Your description makes me wonder if you're leaving the "Formant" knob at its (extremely high) default.  By default, pitch adjustments also result in formant adjustments, which can make things sound "synthetic" or "chipmunky."  If you reset the "formant follows pitch" knob to zero (or something small like 2 or 3), you'll get more realistic results.

RTGraham

sotbee


I get anything but perfect vocals. Even tiny corrections of pitch make the vocals sound horrible. They get this synthetic colouration that is very noticeable and renders it unusable for serious work.

Your description makes me wonder if you're leaving the "Formant" knob at its (extremely high) default.  By default, pitch adjustments also result in formant adjustments, which can make things sound "synthetic" or "chipmunky."  If you reset the "formant follows pitch" knob to zero (or something small like 2 or 3), you'll get more realistic results.

That is correct and I forgot to mention that earlier

In the screenshots I posted above, the formant control was set to zero. No pitch, time or dynamics corrections were applied. I just created the V-Vocal clip and immediately bounced it.

I have it too - Mostly on thin female voices or other subtle things even though tracks are normalized. Sounds like some kind of out of phase...

Here in the Studio the Vari Phrase Technology is called Vari Phase Technology :)

Hi all,i'm sorry to  hear your bad experiences with V Vocal....
It's  incredible weapon for fix my Vox,bass and sax tracks (wonderful for time aligne).
I follow the cake team advices and i ever cut in short clip the track,fix with VV and than bounce to clip.The result is awesome with any click or pitch proplem.
It's only my esperience with VV.

Greetings

Roberto

bitflipper


In the screenshots I posted above, the formant control was set to zero. No pitch, time or dynamics corrections were applied. I just created the V-Vocal clip and immediately bounced it.

Understood.  There's no question that the encoding-and-playback process alone subtly alters the waveform, even without editing any pitches, time, or formants.  Though I haven't tested it, I suspect that the same is true of Melodyne and AutoTune (and I would be surprised if it weren't).  The specific "symptoms" that sotbee described seemed to be related not just to the encoding process, but to the formant control as well, so I thought it might be useful to point it out.

RTGraham

bitflipper


In the screenshots I posted above, the formant control was set to zero. No pitch, time or dynamics corrections were applied. I just created the V-Vocal clip and immediately bounced it.

Understood.  There's no question that the encoding-and-playback process alone subtly alters the waveform, even without editing any pitches, time, or formants.  Though I haven't tested it, I suspect that the same is true of Melodyne and AutoTune (and I would be surprised if it weren't).  The specific "symptoms" that sotbee described seemed to be related not just to the encoding process, but to the formant control as well, so I thought it might be useful to point it out.

I have not actually performed a scientific test on Melodyne, but having moved from V-Vocal for JUST THE REASONS mentioned here, and being very happy with Melodyne,  I'd be VERY surprised if Melodyne does anything to the signal if no processing is done - you certainly can't hear it.

MatsonMusicBox

RTGraham

bitflipper


In the screenshots I posted above, the formant control was set to zero. No pitch, time or dynamics corrections were applied. I just created the V-Vocal clip and immediately bounced it.

Understood.  There's no question that the encoding-and-playback process alone subtly alters the waveform, even without editing any pitches, time, or formants.  Though I haven't tested it, I suspect that the same is true of Melodyne and AutoTune (and I would be surprised if it weren't).  The specific "symptoms" that sotbee described seemed to be related not just to the encoding process, but to the formant control as well, so I thought it might be useful to point it out.

I have not actually performed a scientific test on Melodyne, but having moved from V-Vocal for JUST THE REASONS mentioned here, and being very happy with Melodyne,  I'd be VERY surprised if Melodyne does anything to the signal if no processing is done - you certainly can't hear it.

Again, I haven't specifically tested it, but having thoroughly read the Melodyne manual and their description of how their processing works, I have to assume (hopefully not incorrectly) that even a non-edited note will still not phase-null with the original source audio.  I'm basing this assumption on the fact that it is possible to specifically *turn off* processing for a track, and also possible to select different processing methods.  I think Melodyne's initial encoding mechanism, coupled with their "synthesis" technique of playback and processing, is just more transparent than that of V-Vocal.  Think of it this way: if unedited notes were completely unprocessed in Melodyne, you wouldn't be able to scrub audio at any speed including standstill the way you currently can - *everything* gets processed simply by being loaded into the editor.
 
Like I said, though, I have not yet specifically tested it for phase-null.

There's no question that the encoding-and-playback process alone subtly alters the waveform. . .

Unfortunately, the effect isn't subtle.

In my test, on passages with high-frequency content (i.e. "S"s) the difference wave was almost the same amplitude as the original clips. When soloed, the difference clip has that "metallic" quality that's so often a complaint about V-V.

When I find the time, I will perform the same test on the same clip using Melodyne (Cre8, though, don't have the VST) and compare. If somebody would care to do the null test with the Melodyne VST, that would be helpful.

I'm running 8.3.1.

I have had run-ins with the flanged s's and t's too.
But for the first time last week I was editing a vocal phrase and after tweaking the first part, then the whole part sounded hollow. And this time, not just the s's and t's, but everything. It wasn't wildly obvious so I went with it and so far the client has not spoken of it, but I did a whole bunch of vocal editing under Sonar 7 with only the s and t issue. Not the coloring of the whole clip.

I can testify that this is a new problem though I don't know what version of 8 this coloring started with.

:Ron

I was wondering if this was a new issue too.  It may have always had this phasing issue, but a few vocal clips were recently rendered useless going through V-Vocal.  I don't recall having that severe an issue.

For now, I am just tediously splitting clips down to the syllables that need correction and only applying V-Vocal to the mini-clips, making certain to not splice in any breaths, S's, F's, etc into the clips.

It's not a new issue. V-Vocal hasn't been updated for a while (it's still at version 1.5), and the last update (at S7) added the pitch-to-MIDI feature, with no mention of sound quality.

Based on the spectrum of the null wave (the difference between the original and V-V clips), it appears that the corruption is mostly in the high frequencies. That's why it's most noticeable on S's, which are typically broadband, with significant energy up to 8-10KHz and beyond. Many clips don't suffer any audible damage at all.

V-Vocal is working on my last nerve .  I resigned to tediously splitting clips down to syllables that need a tweak and only applying V-Vocal to the mini-clips.  Now I have pops/clicks everywhere!  I thought it was cpu overload at first, but there is a pop at almost every split that was not done in dead space. 

I was trying to split words so the opening or closing F's, S's, etc were not polluted by V-Vocal.  Crossfades do not work on V-Vocalized clips, so this will probably mean bouncing each clip down and then doing slight crossfades.

The convenience of V-Vocal's Sonar integration is now heavily outweighed by the pains one must go through to assure it does not ruin a vocal.

I guess the snarky recommendation to "use Melodyne" is not so snarky after all!

But can we get some validation from Melodyne users that the investment is worth it, and all of our V-Vocal frustrations are not an issue in that plug-in?
Cheers!

Big thanks for your (and others on this thread) detailed analysis on this - the insight on "how to make it work as best as possible" is invaluable!

bitflipper
Based on the spectrum of the null wave (the difference between the original and V-V clips), it appears that the corruption is mostly in the high frequencies. That's why it's most noticeable on S's, which are typically broadband, with significant energy up to 8-10KHz and beyond. Many clips don't suffer any audible damage at all.

I've been exploring this issue a little more. The data corruption that V-Vocal is imposing appears to be a nonlinear phase shift. As the clip proceeds in time, a phase shift is introduced. By nonlinear I mean that the higher the frequency the greater the shift. Because different frequencies are being shifted at different rates, the spectral content is changing and comb filtering is occuring as harmonics pull away from one another.

The effect seems to worsen with the length of the affected clip. The effect is most noticeable on broadband sounds. This might explain why some clips don't noticeably degenerate.

Below about 500Hz, the effect is almost nonexistent. At 8KHz, it is rather pronounced. V-Vocal appears to roll off frequencies above 10KHz, but this kind of weird distortion isn't likely to be audible that high anyway. It's the critical bands between 1Khz and 8KHz that suffer the most.

The screenshots below show how the waveform morphs over time. First, an image of the test tone I used, which is a 1Khz wave with odd-order harmonics, spectrally similar to a square wave:


The left image is the original test signal. The right image is the clip after converting to a V-Vocal clip and then bouncing back to a normal audio clip. No V-Vocal processing was done. The distortion is obvious.

Here's another view:



The two waveforms on the left show the original (top left) and the V-Vocal clone (bottom left) at the start of the clip. There is no obvious distortion. The two waveforms at the right show the original (top right) and the V-Vocal clone (bottom right). Note the distortion. These are two cycles from the same clip. It's the same unchanging waveform. The only difference is the left-hand image is from a quarter-second into the clip and the right-hand image is from about 1.5 seconds into the 5-second clip.

So, you may be asking: "what if I actually let V-V do some processing?" Will it make the distortion worse?

Here's a comparison after a 1-semitone pitch shift. Of course, there is a severe phase shift, but that's a normal side-effect of pitch-shifting. As long as you don't combine it with the original, you won't hear it. But notice that the waveform distortion did get worse. Specifically, we have some high-frequency attenuation.



EDIT: Oops, I made a mistake in that screen shot. The top wave is not the original test tone, but the first V-Vocal clip - the one without any pitch shifting. The bottom wave is the correct image, though: the original clip after pitch-shifting. The point is still made by the picture despite my mistake.