SL10 workflow for large clips - case study

New SL user here. I decided to dive in on a large project – actually a small project but large files. I had a big band playing at a jazz club earlier this week. I always do some kind of audio recording on the stage. It isn’t for studio quality, but I want material for rehearsal study, and if the recording is good enough, I use it to replace the audio track on the video recording.

On this occasion, the program evolved into an extravaganza with 21 people on a rather small stage, so I decided to go with a minimal recording setup, just because there wasn’t room for much more. What I did was a single stereo mic (Rode NT4) mainly for the horns, and I put splitters on the two house vocal mics – so it was a total of 4-channel recording.

My game plan was to use SL to take the vocals out of the stereo pair, because I had the vocals on separate channels. But then I decided to let SL unmix guitar, piano, bass & drums as long as I was going down the unmix path.

This was a 2-set show. For the first set, I did all the unmixing with SL as a VST in Cubase. The unmix layers end up living inside the Cubase project file, which became enormous, and eventually Cubase became quite slow. Also, I really couldn’t find a good way to get SL removed from the project once I had the unmix done. I copied all the unmixed layers into new Cubase tracks, but this was still connected to SL. I did render-in-place on those tracks, which did disconnect them from SL, but I never really got SL disconnected from the original clips (which were each an hour long.)

That worked, but I don’t recommend it. For the second set, I copied my Cubase project to be the starting point, but deleted all the clips, which got SL out of the Cubase project. I did the unmix within the SL10 stand-alone app and exported all the unmix layers to WAV files. From that point, I simply dragged the WAV files into the Cubase project as if I had used separate mics for each of those WAV files.

That’s a simple workflow. However, it does take a long time to unmix an hour of music (basically the same time whether you use VST or stand-alone). My studio computer is about 8 years old. So far, it has been plenty fast for anything I do in Cubase or any other music software. But I might look at upgrading to do this big SL processing faster. As it was, I just let it run and watched a football game.

How did it work? A lot better than I expected. Without SL, the drums would have covered up everything. In the house, there were several area mics on the horns, so I think the house mix was decent, but on stage, the drums were loud. With this unmixed approach, I had drums-L and drums-R as separate tracks, and they were isolated almost flawlessly, so it was easy to turn the drums down.

Of course, it would be better to have had 20-track recording, but under the circumstances, I’d say this was a really good result. And now that I know it can work with a reasonable workflow (if you don’t mind an hour of unmix time per set), I will probably opt for other simplified recording setups in the future.


That’s a great case study. Thanks for sharing your experience. I just bought SL Pro today and I’m impressed so far. I look forward to remixing old live stereo mixes from my archive.

1 Like

I should mention that my PC has 4 cores, 8 threads. During the unmix process, all 8 “processors” were used, and they all ran around 30-40% busy. The source tracks I unmixed were on a WiFi-attached NAS box. If I had everything on a local SSD, it is possible the CPU would have run near 100% and processed each set in 30 minutes.

Actually, that’s not exactly right. My stereo pair was recorded as 2 separate mono tracks, so I actually had four separate hour-long WAV files, L and R from set 1 and the same from set 2, and each one took about an hour to unmix – a total of 4 hours unmix time.

In a case like that, there may be some logic to getting each set into a single stereo WAV, because SL10’s AI might make some different decisions if it can look at both channels during the unmix process.

I wonder how a newer PC with 12 cores (24 threads) might work. I think in my case, I would be bound by the speed of accusing my NAS box, but I could replicate that material to a local SSD if that made a big difference in unmix time.

1 Like

If you want to speed up processing of unmix significantly, grab yourself a decent gpu.

Yes, expensive. But what a difference.

For “removing” sl from your project. If you are using sl in ARA mode. Go back to the ARA menu and select Make Extension Permanent.

Or simply do you’re Unmix in standalone mode and drag the resulting layers into Cubase / Nuendo.

I have two livestreams that were done about all this if you fancy a skip through:


This bit, he did try… :slightly_smiling_face:

1 Like

Very good, I didn’t have time to read the full essay, sorry.

1 Like

Oh, no apology needed @Phil_Pendlebury - just something I noticed… :smiley:

By the way, thanks for posting your videos; they are always good to follow.! :+1:


Thanks. That menu item didn’t make sense to me. I guess I should have looked at the documentation. I did try the one about removing SL, but it didn’t work.

The ARA mode looks fantastic for small edits. But for a big job, I think I will always use the stand-alone mode.

Bottom line, I was surprised at how useful this overall approach was, particularly considering I wasn’t trying to get studio-quality results. I am thinking SL is already quite close to the maximum that is possible in unmixing, particularly with vocals. As I have given more thought to this, take the case of a simple recording of a band using a stereo pair. There are all sorts of frequencies floating around the room, and many of these cancel each other, never making it into the microphone. Therefore, there is content that is simply not in the file to be unmixed.

Comparing that to the case where each instrument/voice is well isolated and recorded to a separate track, “everything” in the recording. Of course, when you mix, you can get those same cancellations, and that is part of the art of mixing.

1 Like


I’m ready to upgrade my GPU. Uncertain whether to go with Nvidia or AMD. Any ideas which one would help SL processing most efficiently?

Coincidentally, my CPU (still a beast if you ask me) is a 1950x 16/32 core/thread yet any attempt to use SL preview modes results in “CPU speed insufficient”!?! Current GPU is RX580. Same message when I use it to switch to GPU processing. Windows 11. 32G RAM. Samsung 980 Pro 1T. Not certain how to track the bottleneck.

1 Like

Here is a video of highlights from the show I mentioned above.

Again, this all came from a single stereo mic on stage, plus feeds from the two vocal mics in the house PA. I am conducting. The stereo mic is directly behind me.

(It was also a single camera shoot – Canon XA60. The audio from the camera had loads of audience noise, as you would expect.)

1 Like

Great work!