When using component separation and increasing the transient level by only 3db (even less than 3db and certainly worse when adding more) there is a noticeable flanger/chorus (phase misalignment) artifact after exporting the mix-down.
Why is this happening and how can I eliminate it?
This is a deal breaker if I can’t have a clean export after such a simple, subtle adjustment.
Did you apply the Unmix Components module with the mode Reconstruct hidden noise selected? If so, try again with the default mode Non-destructive.
With some quick tests I found that the mode Reconstruct hidden noise may be too aggressive at reconstructing (or more like guessing) missing noise, especially in the lower frequencies (but it also adds quite a lot of higher frequency noise). The result might actually sound not too bad when you only listen to the Noise layer (as the other layers are not changed), but in the full mix its likely to sound somewhat garbled. As Unmix Components is dependent on your set FFT Size, you may get better results with different settings.
I used the default setting of non-destructive.
There should be no big issues with simple gain changes, especially with changes to the Transient layer (of course it is material dependent, but in my limited experience the Transient layer often contains the least amount of the unmixed content).
What does it sound like in SpectraLayers after the unmix process - does it sound as it should or is it already wrong? Did you do any other modifications to the audio before or afterwards? Maybe you accidentally moved a layer a bit?
In any case, the export should sound exactly as the complete composition. You could double check by reimporting the exported file into the project again and make a null test.
Also, which SpectraLayers are we talking about - version 12 Pro? Standalone or ARA extension?
this thread is the reason why to save a safety copy layer of the original and mute it
then user isn’t trying to back track to where user possibly made a mistake
and can actually compare if the unmix module is actually not functioning as expected
I’ve experimented a little bit and yes if I reduce the fft size to a minimum it definitely helps (which makes sense) but I can still hear a little bit of the flanging. It’s most apparent in extremely heavy mixes with lots of cymbals and synthesizers with higher harmonics. I did also notice that when I was trying to add a subtle bit of attack to a string ensemble of only violin, viola, and cello, that there was a bit of this effect as well. I would have assumed this would be more apparent in the noise layer, but just a small increase in the transients seemed to me a subtle adjustment. It wasn’t very apparent when listening through studio monitors but it really stood out when I listened through headphones.
I’ve only recently started using this application and I’ve watched every single YouTube video that I could find. And I do understand fast fourier transforms very well but I admit I was skeptical of how accurate these sorts of modifications in the frequency domain could actually happen. I guess I have mixed feelings but I need to work with it a bit more before coming to a conclusion.
I am though surprised that there are some lacking in the interface and ability to do things like saving vst plug-in settings in the vst module. That really messes up workflow. It’s surprising because in dorico and wavelab that’s not an issue.
Anyway, thank you for the tip about the fft size being important. I will keep that in mind when I do more complicated processes with this application.
Sorry, I missed your question about version. It’s version 12, which I believe is the most recent. The standalone application. Not the plug-in.
Yes, I definitely do that as well.