I’m wondering whether C11’s new vocal extraction feature would be able to extract an instrumental solo part from a concerto. It would be highly useful to create accompaniment backing tracks.
Has anyone tried this?
I doubt it. If this SpetraLayers feature works like Izotope RX, it relies on the ability to recognize the harmonic series characteristic of vocals.
It would be great if AI-based software could eventually identify individual instruments as well as the human ear can.
That’s what I thought, but was hopeful the algorithms were different to RX. AI: give it a few years 
spectral layer 7 pro can do this with some instruments -vocals, piano, bass, drums and other.
I have both SL7 and RX7, and have trialled RX8. I have found SL7 to be far superior and it’s not restricted to just vocal extraction. It depends on the source material – as you can imagine, it works with “vocal-like” sounds as well, and I’ve had success with things like harmonica and flute. Something as dense as a concert orchestra is of course going to be more difficult, but remember it’s (also!) a spectral editor, so you’re not limited to just the automated processes. Given enough time and patience (and skill using the tool) you can achieve incredible things.
You can always ask the Robin (the developer of SpectraLayers) directly over in the SpectraLayers section of the forum.
Thanks all, I might download the trial version of SL7 and see if it achieves my goals. I’m mainly dealing with brass instruments, so whether they’re considered “vocal-like” is debatable!