Thanks for your answer,
yes my bad, i mean rx 8
So i suppose from your response that no improvement where made to the “unmix stem” function in SL 8
SL7/8 and RX8 use the same AI model, so you’ll get similar separation in all cases.
maybe, but definetly, results are NOT similar in my use cases,
maybe SL algo need some more training / mentoring ?
i would be happy to provide some material for you to test if that can help improve the software
Just this morning i was trying to extract a flute from a jazz track with piano / bass / drums, RX does it right, SL got the flutes spreads in vocal & other ( and lighltly in an other layers, but can’t remember right now), and the drums transients where less preserved
i tried 5 differents combination of settings, but can’t get the flute to be properly separated has it was in RX
the bass is usually better separated in RX, attack sound musical and very less artefact from other instruments, it strikes me from my very first tests with the two.
i also have less bleeding in the drum parts in RX ( without tweaking )
in SL i can raise the sensitivity, but usually the point where i get better transients ( usually the low end need this tweaking - kick transients ) is the point where it starts picking up too much of other instruments transients in the high end
For the moment i’m not fluent with SL, so removing those high end artefacts, took me hours when i tried ( for the sake of trying )…
BTW, as requested before, a “preview” feature would really help get the right settings faster ( especially when working with a few minutes long files ) before “unmixing”