Oh you just gotta bounce the Melodyne clips after the transposition and again if artifacts remain try a different Offline Render option (in the Clip Inspector).
 
Try that if you haven't already. Also I would personally  do this with the Melodyne set to Percussive for ease of use (single line of blobs to drag up/down) but maybe popping open the Melodic mode and ensuring all the blobs are accurately detecting the desired note might help the transposition.
 
Are they raspy/growly vox or clear and defined? That might make a difference.
 
However if Melodyne doesn't work try the DSP Transpose as well. It might work better. There is of course V-Vocal if you have it which I have yet to play with.
 
Thing is, at least to me... theoretically, is better to only transpose (not transpose AND speed up) in the digital realm. That's two different actions/processes that need to be calculated (and makes you perform slower for no reason).
 
With tape it's different because it's the same physical data played back/sped up that causes the transposition.
 
But... I could be entirely wrong about that because... well I'm just Beep learning as I go.
 
Good luck.

Craig,
 
Couple of points...
 
1)  I agree that Kontakt isn't the only tool available to play back samples (any sample playback engine will do as long as the GUI has a pitch wheel).
 
2) I only added the tape speed up or slow down effects as added benefits of using a sample playback engine.
 
But wouldn't it be cool if there were a speed dial in SONAR that slowed down the audio (just like tape) but let you record against the slowed down time and then when you speed up the dial, the added track would speed up as well to keep everything in sync and in pitch?
 
Just offering an alternative method to the OP.
 
Jim
 
 

Yes, simple "Varispeed" is one of the highest rated Feature Requests. Other programs have it and I could certainly use it.
 
I think Cake could design it to be better/more useful/task specific.
 
Groove Clips/Looping and the stuff Craig talks about are cool enough for that but it's all multi step processes. A big ole fader knob (with intelligent options) that affect everything in one go would be great.
 
I worry about stability though. Sonar seems to HATE such real time audio mangling from disk. Maybe their current efforts to streamline the guts will make it possible.
 
I'm guessing there are probably a bunch of "Varsipeed" prototype tools in Cakewalk's "Area 51" but they'd cause more mayhem than they're worth... for now.
 
+1 for Varispeed... I want proper shuttle control too.

The problem with a global varispeed is that it will affect audio quality. Whether people hear it or not is something else, but there's no way around creating data that doesn't exist when slowing down, and removing data when speeding up if you want to preserve the duration.
 
FYI Beepster the reason the approach I suggested doesn't have artifacts is that it just plays back the existing data more rapidly or slowly...basically like changing the sample rate.

Anderton
The problem with a global varispeed is that it will affect audio quality. Whether people hear it or not is something else, but there's no way around creating data that doesn't exist when slowing down, and removing data when speeding up if you want to preserve the duration.
 
FYI Beepster the reason the approach I suggested doesn't have artifacts is that it just plays back the existing data more rapidly or slowly...basically like changing the sample rate.

Yeah... the audio quality degradation is a given. The programs that have the feature in a simplified format can't really escape that of course (but maybe their engines/programming/algorhythms can warp better under those conditions... like just good Online algos). I'm just saying having that dial or fader there to do it quickly and globally is convenient and the aftermath can be dealt with later. You know what I mean. We just don't have that "dial" there to make it happen even though all the guts, in theory are there.
 
As for your approach to the OP's question it just seemed more like a logistical difference where if the slow down does not need to occur then why do it if it can be avoided.
 
To me it seems like Time and Pitch are very much separated in Sonar (and the digital format in general) so the slow down is two potentially damaging actions instead of one.
 
I could be entirely wrong on that and if so would love to read more about it.
 
I just figure mangling the mix/backer file (that can be totally thrown away after), recording the track at the lower pitch then raising only THAT file (with no time warping) would be less damaging to the digital data.
 
Essentially the ONLY destructive action being performed is the vocal track being transposed upward. Not sped up and transposed.
 
Seriously... if slowing down as part of pitch transposition in the digital realm is preferable to simple transposition I would ver much like to learn more about it.
 
Cheers, dood.

I started using Craig's method.
The Loop Constructor behaved inconsistently if I dropped the pitch.
I have just posted a fresh thread outlining my experience.

@Beepster: Slowing down pitch to half speed but maintaining the length means having to discard half of the samples.
 
Then, recording my voice at that pitch and doubling the pitch but maintaining the length means "inventing" half of the samples in the new version and guessing what level they should have. At 44k1 sample rate, I will have an audio bandwidth of around 12kHz
 
If the length is allowed to double, all those samples stay in existence. No loss.
And afterwards, when my contribution is pitched up and the length halved, no additional samples have to be "invented".
 
Incidentally, I use Melodyne externally only (I got a full license). The original artist's voice is an alto female voice. Pitched down by -9 steps to match my range makes it sound rough and growly. To get a good ghost vocal, I naturally have to try simulate what I am hearing, then pitch it up again.
 
Using a round number like -9 is indeed better - a half tone up or down makes no big difference to match my register range.

OK... not if you resampled the audio in real time (as if you raised the pitch wheel of the sample player if all the audio was a sample). So, you could raise and lower the pitch and there fore the speed by taking the re sampling approach in real time.
 
Just my thoughts. If there were some resolution that is lost when you speed it up, I believe it would be inaudible by human ears. :-)
 
Jim

Resampling reduces artifacts. But it cannot recreate material that does not exist.
At 44k1Hz, repitching my voice will result in an audio bandwidth of 12kHz max.
For a vocal, not such a big deal. Most dynamic microphones drop off there.
 
With double-length material, all samples exist and are retained, giving 22kHz max.

jpetersen
I started using Craig's method.
The Loop Constructor behaved inconsistently if I dropped the pitch.
I have just posted a fresh thread outlining my experience.

I posted in there...already entered a bug report on the cursor movement, as well as the workaround. However you should have had no issues with the sound, only the cursor tracking.