MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has created software that can take a look a a video of musicians and isolate the sounds of specific instruments to make them louder or softer.
Pretty neat. Right now it appears as if it’s limited to just two different instruments, but I’m looking forward to the time when they will be able to take an audio file and isolate specific players. I want to be able to take a Duke Ellington Orchestra recording, for example, and be able to accurately transcribe the exact voicing that Ellington wrote. One band I perform with regularly will recreate classic traditional jazz recordings and sometimes it’s very difficult to hear specific instruments because of the early recording technology used. Software like this could make it easier to boost the instrument sound that we’re having trouble hearing or turn down the instruments coving up what we’re trying to transcribe.