I’ve just come across this site, not sure if it’s been discussed before: https://splitter.ai
You can upload a song in mp3 for example, and it splits it out into 5 different mp3 files for you: one each for bass, drums, vocals, piano, and other. You can then load them into your favorite DAW and mute or unmute them individually.
I’ve only tried it with one song, but it worked amazingly well. (I guess it helps that RATM has exactly one vocalist, guitarist, bassist and drummer.)
Thanks, I’ll give that a look.
Interesting! Just tried it out on Radiohead’s “Creep” and wasn’t impressed with the results. (I won’t share here cuz copyright blah)
The bass track had a ton of mud from the drums and guitar, which I thought actually made it harder to hear than in the full track, where my ear can contextualize all the sounds it’s hearing into the separate instruments.
The drums and vocals were clearer, but had a bunch of weird sample rate glitchy sounds that were painful to listen to, and some weird squashy volume stuff on the drums too.
Overall - I don’t really think technology will be able to nail this 100%, but I’d say this isn’t quite clean enough to be super useful for transcribing (bass, at least), and probably wouldn’t want to use these tracks for a cover or as samples because of the unpleasant artifacts added by the processing.
I tried The Arc’s “Pistol Made of Bone” The bass track was less than MEH, but voice and drums was not as bad! Thanks for sharing @JoshFossgreen. I’ll try other songs.