Melodyne extension alters track even with no user modifications

As an experiment, I duplicated a track in Cubase and opened one in Melodyne without making any changes at all. I closed Melodyne and bounced the track. I then sent both to a mono output and reversed the polarity of one of them. I expected to hear silence because I had not made any changes in Melodyne, but I heard a clear signal throughout the entire track indicating that just opening the track in Melodyne (i.e., having Melodyne analyze the track) made substantial changes to the track. This is not acceptable. Is there any way to avoid this? Thanks. --Jim

1 Like

Hi and welcome to the forum!

I was all set to write that I cannot reproduce it, at least I couldn’t with a test signal (sine and pink noise), but then I tried it with real world recorded audio, and I can somewhat reproduce it under one condition.
If you take two identical audio events which null completely, and insert Melodyne on one of them without making changes, it doesn’t null anymore (my guess is because Melodyne is just doing some stuff for realtime playback and audition in in ARA).
If you then make the melodyne extension permanent, it reverts back to a null.
You say you did bounce the event with Melodyne on. If you do that, you’ll keep the residue from the realtime mode. My guess is that melodyne (or ARA in general) can have different modes, for realtime playback and for render, and quality modes can differ between those. On render, Melodyne probably notices that there isn’t any change to the audio and thus can pass the orignal audio untouched. On bounce, Melodyne doesn’t get notified for render mode and you bounce with realtime mode.
At least this is my completely uneducated guess.
What you should do then is never bounce audio with ARA extensions on, but always use Audio->Extensions->Make Extension permanent.

Thanks so much for your reply, fese! Your solution does help. somewhat. My concern is that if I make some small modifications in Melodyne (e.g., removing a few unwanted overtones from a guitar track) that Melodyne will also change the parts of the track that I did not want modified in some way. So, I tried this. I removed one note in Melodyne at one location in a track, applied Make Extension permanent, as you suggested, and then ran the mono null test (reverse polarity on one track) against the original track. I was hoping that the only sound I would hear would be the one note I removed but I heard a difference between the two track on the entire track (slight, but audible). I’m just concerned that Melodyne is changing my original track even in places that I don’t want it changed. Perhaps Spectralayers 10 is a better tool for this. Best, --Jim

I guess that to the nature of the processing it is inevitable that changes occur. I somewhat doubt that it’ll be different for similar tools
The question is: do you also hear a problem when you don’t do a null test and listen to the audio in context of the whole arrangement? If not, I wouldn’t bother. If yes, maybe it is not the right tool for task or you’re just processing too much.
If it just one part of an audio event, you could also just cut that part and apply the processing just to that.

Great question about whether it really matters in the context of the mix, which is the ultimate assessment. It just bothers me that Melodyne changes things that I don’t want changed. I think your suggestion to comp in the Melodyne modified parts to the original track is actually perfect for my case because I’m usually only deleting a few unwanted overtones. Thanks again for your help and super suggestions! Very much appreciated.