Yeah, Null test is a low priority, on a truly long list of things to do… Too long.
And that’s probably what Steinberg is thinking with prioritizing the fix for this.
But I will follow this thread…
Please note - I didn’t start analyzing this just for the sake of it but because unlike you I was noticing vocal notes that were out of tune, but only after rendering.
I did null tests to check if the corrected files that had reverted to the original pitch were identical to the original files and in all cases they were except for the first and last 2ms. Performing null tests on files that are audibly different is redundant!
I’m not sure how it can ever hope to work if it can’t at least start by correctly analyzing the pitch, that on its own is clearly a fault and something that Wavelab, their own software, can do absolutely fine. Keeping quiet about a known fault is by definition a con.
I’m not quite sure from your message what you tested, ie. what, where and how you’re pitch shifting, so though your numbers look good (I could live with 5%, no problem!) I’m unclear what exactly they relate too.
I used null testing between files that shouldn’t have sounded the same but did, but it obviously serves no purpose when comparing files that should sound the same but clearly don’t!
Were you using Cubase 13? If you’re getting better results in that I’d be interested to know; Steinberg sidestepped my question as to whether they’d fixed it in 13 but yes, I might well download the trial version to run the same tests.
Of the 4 methods you used, only recording playback (patched out and back in) worked in my tests.
That’s exactly what I thought when I first discovered this - I assumed that it was my hearing, or what I wanted to hear, that was the issue, or possibly even what was surrounding the vocals, as I had complete confidence that not only was what I was seeing on the screen correct but that it would render files how they had been edited!
I did wonder whether it was coincidence that Steinberg added Melodyne Essentials to Cubase in the middle of my interactions with them in 2022, and I also wondered if it would be a solution, but the size of the additional files it creates just made it impractical for me
And you’d think they’d want to, but apparently their professional pride doesn’t stretch that far
1 Like
Basically I tested if creating an audio file that uses VariAudio results in something different than real-time playback. Any difference was neglectably small.
But I don’t have your audio material and, yes, I was testing it on the latest version.
If you could check if a newer version of Cubase makes a difference or if you could provide a small audio sample (a few seconds only; I know this is not always possible if working in a professional context) that would be swell.
A null test producing silence is not disputable. If an inverted signal cancels out another signal both signals are identical.
But any non-silent result can lead the developers to the road of making the algorithm better. (Here we are at sharing a sample, again.)
Perhaps I’m misunderstanding something in this discussion, but if I take a sine wave with a note pitch of G4 (≈392Hz) and VariAudio tells me it is actually of a slightly different pitch, then that is an issue you can’t prove with a null test.
I’m happy to supply any sample you’d like if you’re specific about what you need. Re. the developers - I sent everything I was asked for, and you don’t need a null test to compare two files that are 40 cents apart!
And, as I said in the original post - they admitted that there was a problem that they hadn’t yet found a solution for.
The post wasn’t about whether there was a problem or not - THERE IS! - it was about why they can’t or can’t be bothered to find a solution for it, and why they think it’s OK to advertise it as if there isn’t.
Don’t worry - I think the null test thing is a bit of a red herring
Awesome. Is there a chance to keep each sample to a few seconds or do the audio examples need to be longer? Feel free to PM me the files or the download link. I will run it on my install and have a listen.
Oh , I don’t know about this kind of stance on Steinberg.
In all honesty, I waited from CB11 to jump to 13, mostly because I was too busy to change, but also because the issues outlined here on this forum put the fear of God in me 
But, when I finally took the plunge, I was extremely impressed and happy with the amount and quality of the upgrades, including things that should have been done about 20 years ago such as simplified Logical Editor, Mono to Stereo conversion on a track, and my pet hate, placing the freaking blue line on where you wanted to record at Step Recording (!!!@* 
) among many fantastic upgrades.
Plus they gave away the Iconica Orchestra and a lovely felt piano, and also effects, that are as good, if not better than 3rd party plugins.
So, I think they haven’t addressed the Variaudio issue because they are working on other functions.
1 Like
Has anyone done a test between Melodyne and VariAudio?
The same files with the same corrections and then compared the results?
That way you an tell if it’s a Steinberg thing or perhaps a “this is how the tech works” kind of thing.
No problem, but you need to be specific about what you want me to send - do you want me to repeat the Retrologue test?
Fixing the problems they have would seem more important than new features , especially when it’s something as utterly fundamental as whether things render correctly, whether that’s VariAudio, fine tune or anything else that might be affected
It’s ok. I thought you have some vocal recording or so.
I will try to dive a little bit deeper into the matter by using Retrolugue audio.
I haven’t but this is not a ‘how the tech works thing’, it’s not even a Steinberg thing, it’s a Cubase thing. Wavelab, even Wavelab Elements, can analyze, pitch shift and render files seemingly without any of these issues
Okay, I have ran your scenario. Changing pitches with VariAudio, then bouncing, RIPing, mixdowning, manually recording this event.
And I keep getting the same result over and over again. There is no audible difference.
As a source I used Retrologue with one oscillator, sawtooth, fixed phase at 0° (if it is free phase there will always be a difference if you compare the VST track with a rendered file).
So maybe it depends on the source material in order to replicate your issue. Please send me either a wave file or a Cubase project together with instructions exactly what I should do in regards to VariAudio.
2 Likes
Wow, so you did the test with the same material in Wavelab and Wavelabs results were better?
Weird as you think that both programs would use the same audio engine and code base.
Tho Wavelab is pretty wacky so who knows
??
Wavelab is developed and supported by Philippe Goutier, entirely seperately from Cubase / Nuendo. I’m not aware of any codebase or audio engine sharing (although obviously they all support ASIO and VST).
And Wavelab is most certainly not wacky. It remains one of the premier digital mastering tools in the industry.
Other than that, you’re spot on.
4 Likes
VariAudio is the handiest vocal correction tool I have ever used. Period.
5 Likes