Using Unmix Components Twice for More Accurate Layers


I’m just playing around with the Layers Features to learn more about them. I did an Unmix Components on some audio and noticed that in the resulting Noise Layer there are still some Tonal/Transient sounds. So I selected that Noise Layer and ran an Unmix Components on that. Together all the resulting layers still seem to produce a non-destructive version of the original audio, and it also seems as if my New Noise Layer, that was produced from the Unmix Components on the Initial Noise Layer, has less Tonal/Transient sounds than in the Original Noise Layer. So by merging the First and Second Tonal Layers and the First and Second Transient Layers, I’m getting better sound splitting than on my first pass through the audio.

I was just wondering if that makes sense based on how the software and AI is processing the audio?



That could make sense indeed.

Just love :smiley: when there are ways to help focus/fine tune/experiment with algorithm parameters… Or in this case, when there are useful ways of applying, choosing area or lets say “tooling” around the algorithm themselves, which probably defines what Spectralayers has become good at along the years.
Quite the opposite to the one-knob, a few settings - if any - multifunction plugins.

So now, a collaborative community sharing usage tips, ideas, feature requests and feedback, is great news! Thanks Mike.

Nspace - you’re welcome - just tinkering around.

Robin, when doing an Unmix Components, there is an option for Reconstruct Hidden Noise. The documentation says “When Reconstruct Hidden Noise is checked, noise hidden under tonals or transients will be reconstructed.”

Can you explain what that is actually doing?



Rather than explain with words, let me show you :slight_smile:

Here’s an example with noise and tonals:


If I just show the noise layer, see the parts that overlapped with tonals don’t have any data:

And here’s the noise layer when Reconstruct Hidden Noise was checked:

That’s useful when you’re specially interested in the noise layer, and don’t care if tonals+transients+noise doesn’t not exactly reconstruct your original signal.

Thanks Robin - I guess a picture is worth 1000 words - this really helps.

Hi Robin,

Great pictures to show what the Reconstruct Hidden Noise feature does. :slight_smile:

I don’t have Spectralayers Pro 7 yet (still deciding whether to press the button) - I’m still on Pro 5, but often after removing an unwanted sound (like some wrong notes) I can still kind of “hear” a “negative” of the removed sound in the remaining mix (and like your third picture) still see the gap.

Can this (new*?) Reconstruct Hidden Noise feature be used with music to lose the “still hear a negative of the removed sound” that I experience? Or does it only work on noise?

Can it also be used with the auto Unmix feature?

(* Is Reconstruct Hidden Noise new to Pro 7, or Pro 6, or do I actually have it on Pro 5?)

Hi Robbie,

The Reconstruct Hidden Noise is new to v7 (it’s part of the new Unmix features).

However what you might be looking for in your case is Heal (instead of Delete). SL5 has an initial Heal implementation, but it was much improved with SL7.

Thank you Robin. I’ve never tried the Heal function. Infact as only an occasional user, I’ve never really fully got to grips with SL5, having felt a bit overwhelmed after upgrading from SL2.