Unmix is a fantastic feature!

Three weeks ago I received from a friend a 43 yrs old stereo master tape, from a radio show my band played in, recorded at Sveriges Radio in 1977.
The music style was instrumental in the vein of Osibisa-Santana-Weather Report-jazz rock whatever. The instruments were drums and percussion, electric bass, electric guitar (me) hard left and tenor saxophone hard right. The other guys were 25 and I was 17.

A friend baked the tape and digitized it.
It got rinsed from wows & flutters in Celemony Capstan.

I got the 24 bit stereo file and started to try to ”remaster” it. There was quite an imbalance in the stereo field, with the left channel sounding much stronger than the right. Easy to fix in WaveLab 10. But the sound all in all was weak and anemic with weak drums and and an obviously far too close miked guitar with unpleasant details in the upper register.

Upgraded to SL7 a week ago.
Performed the ”Unmix” process and beefed up the resulting drums/percussion stem with VST 3 plugins (Pro-Q3 and some compressor/maximizer).
The bass stem was further decomposed into it’s three components (tone, noise, transients), which were eq-ed and compressed separately.

There I got a tight rhythm section with a more heavy and modern sound, nice!

The guitar and sax was panned very hard left & right in the resulting ”Other” stem. Now, if I could separate them into two mono files, or a split stereo stem, it would have been great.
I haven’t found a function for this in SL7, so I duplicated the layer and applied to each a VST-plugin that could pan each layer left and right respectively,
Then I turned the phase for one of these two layers (to clean out possible remnants from the other layer), and from there I could EQ them separately. I could also erase a very embarassing error on the guitar :blush: At last, after 43 years!
Panning the instruments back, not so hard though.
I then merged them together and applied a little reverb with a VST-plugin.

Voilá, a completely (almost, drums and percussion not separated from each other) new and fresh remix of the tape that was recorded when I was 17 in 1977. Feels good, the result was stunning!

One remark: since SL7 can’t unmix the guitar from the sax, it would have been of great help if there had been a pan control for the individual layers, e.g. next to the volume slider.

Of course I could have exported all stems to WaveLab or Nuendo but I chose to do it all in SpectraLayers.

Thanks for an awesome program!

Travel 43 years back in time, grab a recording and bring it back to 2020! What a great success story – don’t forget to let us know where we can hear the results (perhaps a short clip over in the Made with Steinberg section?) – or “Re-made with Steinberg” :slight_smile:

I have yet to upgrade to SL7 and am looking forward to the advances. The most recent job was done using the unmix function in RX7, where it’s called “Music Rebalance”, and subsequently cleaned up in SL6. It’ll be great to be able to do everything in SL7. BTW, shouldn’t we be calling this process “de-mix”?

I have done a number of these types of jobs over the years, usually from cassette, if I’m lucky from 4-track “portastudio” machines where e.g. only keyboards, bass and drums were mixed down to a stereo pair, and the results can be astonishing. It is very dependent on the source material though, and as you will have seen, vocal type material is difficult to unmix from certain instruments because of the overlapping frequency ranges.

I think it’s important to preserve all the intermediate material as part of your archive. I revisited a recording I had done some restoration on 15 years ago, and at that time, everybody was happy with the result. This time I was able to extract the vocal from a track with acoustic guitar which had been recorded with the same mic – not easy, but possible, which was not the case 15 years ago. Importantly, the best results were obtained using the original “raw” digital file taken straight from the tape, which I had luckily preserved. Being able to recover a great vocal track from a musician who has since passed was very special.

Oh! I want to hear a comparison of that!

Marshall, a whole Youtube video of your process would be so great since SPL 7 is a visual sound processing software. SPL 7 technology is so advanced and new, I’m sure many users needs this kind of tutorial.


Yeah, it’s great fun playing with SpectraLayers.
It will crash or hang every now and then though, but I hope the update of next week will fix most of those issues. It could also be my iMac (i5) being too weak.

Joss, that would be something for Greg Ondo or the other pro guys to put together. I think it is a straight forward procedure to play with the unmixed stems.

MrTopo & MrSoundman, today is the last day of this year’s summer holiday, back to work tomorrow, meaning there will be only little time for fixing a listenable mix - I’m still testing different approaches in the quest for the best sound; the one described is only one of many ways.
I mean, it will sound differently if I change the order of doing things. There could be less or more remnants or leakings ”from one stem onto another” depending on the original material. Should I clean this up or not? Would that introduce phase artifacts?

It is easy to get (comb filter) artifacts on some stems and I’m experimenting with how to reduce them. Would it for instants be better or worse to compress or maximize the original file before opening in SL7?
I believe the source file should as untampered as possible, but I haven’t tried the different variants yet.

Today I found the ’Process › Amplitude › Channels Remixer…​’ which pretty much seems to deal with the panning I mentioned in my first post. I will try this instead of the VST panning plug.
The quest goes on!

nice post - thanks for sharing Marshall :slight_smile:

I just upgraded to SLP7 and have run unmix on some old material … and I am picking my jaw up off the floor!

On the job I had previously done using RX7’s Music Rebalance with subsequent cleaning of artifacts in SLP6 (about 30 hours work), the entire file was processed in SLP7 in 20 seconds and delivered a far superior result. I am simply blown away.

Robin: respect! :sunglasses: