… yes, it’s fast and it’s neat to have the resulting audio being arranged to separate tracks.
But what about the audio quality? I think it’s not even up to par with that of the fastest sep mode in SpectraLayers 12, or earlier versions of SL.
What’s the general opinion here? It would have been nice to be able to use the internal stem sep in Cubase 15 to aquire drums or bass from a track, without going the SL ARA way, but the audio quality of the resulting stems simply isn’t usable.
Right now, SL12 sells nicely because of the HQ stem separation. They wouldn’t want to cannibalize that… 
It’s a great addition to CB and super useful to have for certain quick tasks. But I compared it with a current project where I had already separated a track both with lalal.ai and RX11 and the CB stem vocals have a huge amount of bleed from instruments where you hear none of that with the other tools. That makes it unusable for some mixing tasks where I have had to turn to separation in the past. I’d gladly sacrifice speed for a bit more quality.
For the sake of performance and they probably are not relying on the world’s favorite proprietary compute framework (CUDA) because some people might not have it, it’s probably some CPU algo using a small dataset. Spectralayers Go already adds almost 5 GB to the Cubase package.
I think it’s nice to have it integrated but I doubt I will be using it given that I have SpectraLayers 12. A bit of a lead balloon tbh!
Fast & Crash - macOS Tahoe
If I were to use that feature, it would be to then go from audio-to-MIDI, after the stem separation is complete.
Same here. One nice aspect is that some percentage of users who use the internal 4-part demixer enough….may at some point graduate up to Spectralayers Pro when realizing they need more capability.
fast until crash on a Apple MBP 16 M4 64 1000 40 and no i did not update to tahoe. Still the last OS before tahoe. Please dont tell me tahoe is a must …