I’ve got a problem with SL Unmix Drums process. I’m unmixing a track comprised of programmed drums and a single fast sequencer line which was recorded at 96kHz. Step 1: Unmix song… no problem. All good. Splits nicely into drums and sequencer. Step 2: Unmix drums… problem. Random added spikes, mostly in the bassier drums. What to do? Tried doing this process several times with this and similar tracks and it always adds spikes. (Same with SL12.0.10 and 12.0.20)
Screenshots of the same section: top - Drums after Unmix song, bottom - after Unmix drums
It seems to happen at both the sample rates I’ve tried this on - 96k and 44.1k. There are no spikes in the original file, or on the drum layer after the Unmix Song operation. They only occur after I apply Unmix Drums to the drum layer. The spikes are not stereo and appear in several layers at various strengths.
I thought at first it might be electrical interference so I re-ran both operations on one file and the spikes were all in the same place second time around. They appear whether I’m using the computer at the same time, or whether I leave it solely to Spectralayers. I’m using a MacBook Pro with the M1 chip with 32GB of RAM on Monterey OS 12.7.6.
So after erasing the spikes by hand in Spectralayers, using the eraser, I have found that it is adding extra spikes when exporting the layers, which I can see when I open them in Cubase. This is getting seriously weird…
I don’t understand what you’re driving at. I see the spike, I erase it. I check with my eyes and ears that the click has gone. You can see from original screengrabs what was added.
This thread isn’t about the efficiency of SL at unmixing whichever source, it’s about the fact it is adding noise spikes at random where none existed when using the Unmix Drums module.
You tried on how many different sources? One type of drum programming? Or multiple types of sources?
I have had issues unmixing programmed drums in SL…sound sources that are electronic in nature are not recognized the same as organic sources IME. You really need to provide more information or test more sources. I have unmix song and unmix drums on several sources: real drums mixed down to stereo with other instruments (full mix, “demo/ practice room recording”), live drums of real drums recorded from FOH to DAT and various programmed drums: factory drums within a synth patch, ie HR16 from years ago, live roland octapad. I, personally, have not experienced any unwanted spikes in my unmixing of song, nor unmix drums.
If you would be so kind, can you share with the forum what sources have you specifically tried? I’m not asking for the actual audio, just a list.
My suggestion has some bearing on efficiency, but ultimately, using the eraser tool is a destructive process, period. If you want to go back, the only way is work off of a safety. You cannot show yourself the difference between what you have erased and not by anything but your own memory. Or comparing to the original (as I already mentioned). I’m driving at helping others be able to realize a trail of their work.
It appears you feel I’m not trying to help, on the contrary. Some sources just unmix better than others. What works best is a moving target. If you shared whatever you are using as a source, that might help yourself and others.
The originals were mixes I ripped from an old CD and cassette. Just trying to unmix (using SL, not via ARA in Cubase), partly to separate the tracks enough to put them through Melodyne and get some MIDI info (the original Atari floppies and files don’t seem to work through Steinberg’s suggested conversion methods, but that’s something I’ve addressed in another thread), and partly to re-work some old archive material. Not bothered about having a trail of work, especially as Robin seems to have fixed the problem, so when the patch becomes available I’ll redo everything anyway.
I didn’t mean to suggest you weren’t trying to help, but I was focused on trying to troubleshoot the issue I was facing, and your good experiences didn’t really shed any light on that.
I’m still getting to grips with SL and have a lot to learn, so I didn’t know if I had actually made some procedural mistake with the way I went about the unmixing. For example, today I learnt that Level Unmix was a lot better at unmixing certain sounds than Unmix Instrument. I know SL has its limitations, but you can still almost trick it into giving you what you want by using a non-obvious process. There’s precious little in the way of videos which delve into this on the net - yes, plenty of basics videos, but I haven’t found any which show hacks / tips which are off the beaten track, so to speak.
I have a colleague who uses unmix crowd noise for all kinds of NR
So the sources were major releases? or old recordings of yours? Still not clear if those programmed drums are samples of something ? Drum machine? No worries
Back in the 90s I made an album of instrumentals which, in the end, I didn’t release, as releasing a CD was too expensive and obviously, there was no internet to release online. So I just burnt myself a CD, which I’m revisiting with the idea of a release. A lot of it holds up but I want to replace some of the more dated elements and give it a general spruce up. Also I want to extract the MIDI info so I can think about playing some of it live… I think the drums were from my Casio FZ1 sampler.
The cassette was an earlier project using a Roland SH101 and, I think, a TR707 drum machine. But the 4 track cassette quality is fantastic, so another re-work to do! The only thing was drums and sequencer were on the same track to save space for other synths on the other tracks, so I’m just separating out the various elements, again for MIDI extraction etc.
It’s great that Spectralayers gives us the chance to do things like this.
I’m amazed at the quality of all the TDK SA tapes I used to use back in the day…holding up so much better than the $1000 worth of DAT tapes. I hear you, on these, I have a fair amount of this type of material as well
I was just doing the same thing. SLP12 drum separation in unmix song and drums is pretty impressive in my tests so far. The one I just did was Kurzweil K2500 drums and organ and the results at High Quality are pretty good. The future is bright with SLP in our lives IMO
thanks for that