brassplyer
Well-known member
There are modern AI tools that do a pretty good job of separating instruments on album tracks, do they work on live recordings, say of a jazz combo and a singer using the onboard mic of a camera?
Will AI audio tools separate instruments in live recordings?
- Thread starter brassplyer
- Start date
rob aylestone
Moderator
They only do a good job on certain instruments that differ from others (at the moment, next month? Who knows?)
I've got Spectral Layers and it can separate voice, bass and guitar - and deal with percussion, but it cannot pull two guitars apart. If your jazz combo is bass, drums, keys or guitar and the voice then that will work. The degree of separation is influenced by room acoustics and clapping - audience noise can be difficult and it seems to mess up the higher frequencies when it is removed. It also seems to work better when things are panned across the soundstage. I guess it helps the separation. Distant mics are much worse than close miked sources. I tried a bit with a choir and a piano in a church from an iphone. Recording quality was pretty good, but it struggled separating the voices from the piano - it did it, but the piano was sort of 'warbly'? Not successful.
I've got Spectral Layers and it can separate voice, bass and guitar - and deal with percussion, but it cannot pull two guitars apart. If your jazz combo is bass, drums, keys or guitar and the voice then that will work. The degree of separation is influenced by room acoustics and clapping - audience noise can be difficult and it seems to mess up the higher frequencies when it is removed. It also seems to work better when things are panned across the soundstage. I guess it helps the separation. Distant mics are much worse than close miked sources. I tried a bit with a choir and a piano in a church from an iphone. Recording quality was pretty good, but it struggled separating the voices from the piano - it did it, but the piano was sort of 'warbly'? Not successful.