Can't even separate a metronome from piano

I am finding that this doesn’t separate very well at all. I have a track where piano is played with a metronome and I wanted to remove the latter and used the module to separate piano and drums and it’s still on the piano track. Another project had piano and drums and it didn’t separate those either. Huge disappointment, especially when someone just told me they used Logic to get a midi version of piano including chords, which I also do not find cubase capable of.

“This”? More information required

You’re sort of all over the place with topics. Going back to your first complaint, ok…how well does Logic demix a metronome from a piano?

Which version of Spectralayers Pro do you own …or are you on the trial? Which commands have you tried?

Well I just thought that if it could separate chords notes of a piano and identify them, separating from a metronome would be no challenge. I see that I used Pro 11, and am using it within Cubase Pro 15. Should that be a problem? Instructions said to separate metronome as a drum.

I see I bought and downloaded 12 in December but that I haven’t gotten around to using it. I’ll see if that is better. I will have to refind the song as it was awhile back and when I twice went to post to this forum, the text already there was overlaid on the area where I was supposed to type my problem and I couldn’t. As I said it was within cubase and I’ve worked on so many other songs since that it isn’t showing in the list of recent.

Looking further, I found the project, but since it was within Cubase, the extension doesn’t say what version of Spectralayers. So I am now opening 12 and importing the audio file, unmixing song at high level with piano and drums selected. Since I mainly would use Spectralayers in Cubase, how do I know which version that is using? Since I have regularly been having disappointment with spectralyers, I may not use for months, so recollections aren’t memorable.

Have you tried using “select similar” instead of going through the unmixing route?

Here’s how Spectralayers 12 did with piano and metronome.

Drums, only heard a little sound at about 12 minutes:

Green is piano, only about half of it.

Blue is Other, but actually most of the piano and the metronome. Terrible job in my opinion.

Agreed.

@robirdman1
Which version of Spectralayers are you using?
What is the resolution of the audio you are processing?
What kind of piano are you referring to?
What kind of metronome sound are you referring to?

Please upload a short extract of the audio you are processing so others can test for comparative/confirmation purposes.

why?
Do you think the training materials used for free AI stem models (as used in SL) include metronomes?
I would use spectral editing to remove each metronome sound, not an automated stem extraction. Do each click, one by one, you will derive a perfect result. It takes longer, but *it works
*
I’ve done this countless times with headphone click tracks audible in vocals.

Agreed, select similar should work well…might take a little while for SL to search…possibly similar to unmix song timings

And user has to be willing to use selection tools to create a selection “tight” enough of the unwanted sound to feed into select similar. Just making a quick time selection and hoping SL “knows” what the user wants to separate is not going to produce a worthwhile result.

Yes, I tried to separate a click from a vocal in SL multiple times and it always failed. Went to Logic and the click and vocal separated beautifully the first time.

In S Pro 12 manual it has this link to remove click from guitar and metronome. While mine is piano, I though should work likewise. I tried to export a sample as requested. .wav not allowed. I tried mpeg layer 3 and it uploaded but I only hear extreme faint sound. So how to provide?

using what process?

The only process I know in SL - separate stems. But then I tried separate chorus - nope. Then I tried treating the click as noise and tried to get rid of that - nope. SL is amazing software but at the present time for separating stems Logic and LaLaL beats it every time. No ghost voices, weird phasing issues, voices dropping out or so low as to be indistinguishable, etc.

Hi @robirdman1, to upload a short wave file, zip the file and post that here. Zip files can normally be uploaded without issue, if the file is not too big.

piano w/metronome sample.

piano idea test.zip (1.7 MB)

Thanks for uploading.

TBH I don’t understand why you would call the percussive sound a metronome. The recording in general seems damaged. Are you sure this is the sound you meant to upload? or is this somehow a mistake? The file is stereo but appears to have been recorded on the left channel only.

Spectralayers can do miraculous things but in my humble opinion you might be expecting too much in this particular case. It’s not comparable to the video example you cited above.

The closest you could probably get to separating the percussive sound from the piano sound would be to use ‘Unmix Components’ and then solo (or mute) Tonal.

I think the metronome is clearly audible and therefore some AI should be able to filter this out. But, if SL has a proper modell to handle this yet, is indeed doubtful.

Sorry, it might be beside the point but that sound is not what I would associate with a metronome… not that it really matters! But calling it a metronome seems rather misleading.

Yes, but sticking to what is available in SpectraLayers, ‘Unmix Components’ (as suggested above) gets close to a reasonable result.

Alternatively:

Just short segment of an old home recording, not studio, and sometimes on only one of the channels. It is a polyrhythmic metronome, the Trinome, so there may be 3 different sounds on it (click, clack, bell). It is easy for me, and other humans, to distinguish the metronome from the piano, so perhaps AI, if used isn’t really very advanced here yet.

Which is literally not a specific SLP process. The term “stem” is mentioned once in the owners manual and the sytax “stems” is mentioned 10 times referring to Unmix Song and Unmix Drums modules. So you know how to run modules on source material in SL, ok, thanks, that is much clearer.

There are many ways to attempt to arrive at a desired result. Obviously other softwares produce better results for music stem separation. So you don’t have any experience manually selecting and isolating sounds in SL, you only ever run modules?