Re: [MLUG - DISCUSSION] so this is pretty scary [MUSIC]

["Jonathan King" <EMAIL:PROTECTED>]

On Wed, Mar 12, 2008 at 10:41 PM, Vern Green <EMAIL:PROTECTED> wrote:
> This is not scary to me, this is downright exciting if it works as well as
> they let on which I am skeptical.

My big fear is that you could easily create stuff that no human could
play in person. It wouldn't be less musical, but it would make live
performances seem less special. I'm not sure that's a good thing.

> I doubt it would work well on any mixed tracks.

Arguably, a track with a guitar chord is a mixed track that is kind of
worst case in that all of the pitches are very similar in timbre.

> For instance a track that
> has guitar, sax and bass on it, along with a piano, drum and vocal track,
> but even if you could only do this with a single track it is pretty awesome.

OK, so the usual requirement for separating mixtures (tracks) with N
different signals (instruments) on them is that you need at least N
different copies of the mixture that vary in the "volume" of the
signals. As it turns out, you can get away with fewer IF you can make
assumptions about the kinds of signals (instruments) you are
separating.

I think separating piano chords in some cases could be tougher because
of the similar timbre and the huge effect that the sound board and
resonances from other strings have on the acoustic piano. The result
could sound pretty odd if you "mis-assigned" the harmonics to specific
notes in this case. I think it is telling that they didn't just
demonstrate this on piano.

jking

_______________________________________________
discussion mailing list
EMAIL:PROTECTED
http://mlug.missouri.edu/mailman/listinfo/discussion