- Analyse 15 seconds of audio in VariAudio
- Raw values:
E0 -20%
G0 -3%
A0 -9% - Adjust to 90% tuned. Adjusted values :
E0 -2%
G0 -0%
A0 -1% - Analysed values after using either Flatten, Bounce or Export:
E0 -5%
G0 -1%
A0 -+1%
Is this the expected behaviour?
1 Like
Iāve subjectively noticed flattening sometimes seems to change the VA results, but never quantified it like the OP did.
Anyone here know why the OP might have gotten these results?
I have tested this rigorously and I was surprised by the results. Strangely, no-one seems bothered. I keep having to pay mix engineers to re-tune vocals and I prefer the Cubase results - but they donāt export well enough for them.
So youāre re-analysising already processed audio and then relying on the analysis results, is that correct?
As the final processed audio is not going to be perfect iād say thereās always a chance of the analysis being thrown out when re-analysed again. But must admit, never tried it.
Have you tried routing the output of the live variaudio track into a new audio track and recording it live? Then compare that to the flatten/render to see if it nulls? As that would be my first question as to whether the render is accurate or not.
Also, are you using sine tones to test this, or actual harmonic content such as vocals or some other solo instrument? I may have a play later to see what you mean.
Iām using real vocals. Itās driving me nuts. I only re-analysed it because I was listening to it soloed and thought āthatās odd - I fixed that noteā I look and itās another value altogether. I wish Steinberg would respond with - a) this is how you best achieve a rendered vocal with VariAudio and b) hereās the margin for error. This is a real problem for me. I canāt send this song without working it out. Or the engineer will charge me for Melodyne and I donāt use blanket autotune. I spend hours on this stuff. I just want what I hear, to be rendered correctly. No-one I know who is any good, mixes in Cubase. This is not a criticism - itās fact. Seriously, Iād bounce it through an analog channel but it adds noise. Iām wondering how this test would go in Nuendo? Is it the same algorithm?
My guess is that the value you see after adjusting, is the original analysis of the audio multiplied by some factor, rather than a re-analysis of the output audio.
Thus, it makes total sense that a new analysis would show a different value, because all this pitch correction software is a bunch of heuristics ā meaning, itās not a mathematically linear function, and re-running the analysis on processed audio will come up with a different result.
If 90% correction doesnāt do it for you, try 98%?
1 Like
just be done with it and use melodyne plugā¦sorry thats my advice
or better still fight the perfect culture and reintroduce some human imperfectionā¦just rehearse and spend the time getting it right on the instrumentā¦when I track someoneā¦more than 3 issues and its a retake. Comping has its on issues
I think Iāll take your advice.
Sorry - we are in different universes here. Iām not seeking perfection. I donāt make anything perfect. If I buy a camera and it doesnāt reproduce JPEGS like it does to a screen, I buy another camera. This is a fail from Steinberg, on a process they sell as a feature.
I didnt mean it negatively. I think fixing a couple of mistakes in a performance is no issueā¦but after doing pitch correcting for about 10 years straight on various stuffā¦you cant unnotice now hehe.
Has really cut down my reference track library though.
Time alignment is ok in Cubaseā¦but for the amount I now use pitch correction; I already had melodyne and its not that expensive tbh
It would be nice if there was a succinct vid from Steiny though demoing an actual proof of concept in real world applicationā¦would stop a lot of neg as you def know it can be done
Luv to hear some of your work 
1 Like
@pkmusic i have used variaudio for years and could never get the render/bounce/flatten version to sound the same.
1 Like
Hmmm. This is what I was afraid of. OK. I have to move on. Iāll buy the basic Melodyne. Thanks.
I appreciate your time in posting, sorry if I came across (as cross) Iām just disappointed I have to do all that work again. These are backing vocals and thereās so many. All I do is fix the strays but Lord itās time consuming.
Melodyne is top of the heapā¦reallyā¦its money well spent. No package can do everything well and if its something you spend a lot of time on then you wont be sorry. The upgrade does polyphonic as well which is great if its a studio that does live takes eg guitar and vocals at same time.
Really, speed is everythingā¦if you are spending that much timeā¦you get lost in the forest. Im speaking from experience as I use to spend so much time on edit for my own stuff. I ended up just binning it and doing the takes again but with a lot of rehearsal and it just improves your artistry on the real stuff; it really does.
These days I do mostly only live artists and song can be recorded and mixed in 1-2 days. ā¦and its a real joy tbh
Hope it all goes well for you.
1 Like
I feel vocal tuning is more of a production choice, than mixing. And that should be where Cubase is strong.
Itās odd that you correct a note and then itās not correct in the render. Iāve not come across that before. Iām definitely going to try it out this week though, as it has me curious.
Personally, for vocals, I prefer the results that variaudio gives than I do melodyne.
1 Like
Thank you.
Firstly, thanks for your help with this. The problem (as jwatte said) appears to be in the way the analysis functions. However, the processes for export/render/bounce seem confusing to me. But hereās the wrap up, using invert phase.
- Render. This appears to be successful. The two tracks, phase inverted, do cancel each other out.
- Export. The exported track plays, but very quietly.
- Bounce. Itās as loud as the original. So nothing is actually done except combining audio segments. This may well be intentional, depending on your interpretation of ābounceā
I love Cubase and theyāve made huge improvements to the export function. But this seems like something that could be further improved. And out of fairness to the software engineers, I should say that I was wrong. VariAudio, using the Render function does work; even if an analysis cannot give the same result.
1 Like
FWIW, your headline can now seem misleading and may cause others to avoid using it.
I use VariAudio and have Melodyne Assistant as well. I much prefer VariAudio because it it easier to use and seems to be top of the line without paying a top of the line price for it. (Melodyne costs as much as Cubase or more at full price.)
Moreover, I did some reading in various forums and found people commenting about how the polyphonic pitch correction is unreliable and can leave artifacts. So thereās a hype storm around the selling of that feature, too; one that seems not to prove itself in real world usage.
Reason has a good pitch correction feature, surprisingly. But VariAudio is what I now use for everything.
Im a long time user of meloā¦but Iim over relearning stuff too so it might be biased inforā¦you can get sales pretty cheap though. I have used the polyphonic and its like anythingā¦it will vary with context but for basic stuff like vocal + guitar with bleed issuesā¦its pretty decent. The fact that there is no reference other than 440 in variaudio makes it useless to meā¦I havent worked in 440 for a long time and as mentionedā¦it is very rare for me to pitch correct, its not part of mixing, I just give them other contacts if thats what they want.
But Im glad you got a solution and all the best with it
If render is always required to commit VariAudio (and the result does not return a visual confirmation on the success of this method), it adds an extra step to the task of sending out stems. If a track is only 4 bars long, it needs rendering and then exporting, so as to be the correct length. Rendering should be included in the audio export function. It would certainly save some time.