Slashdot Mirror
Simulating Human Musical Performance
GFD writes "The EETimes has a story about a software program that can mimic the subtlties that humans give to a musical performance to give it a particular emotion or style. Apparently it is already so convincing that some TV producers are using it rather than live musicians. Long range plans are to clone the musical styles of famous musicians. Interesting question - there is no question that a Jimmy Hendrix could copyright his music but could he copyright his style? " Bet they can't mimic all 3 of my power chords.
"now all pop music can be created by formulaic computer programs!"
."
Better than being created by formulaic record producers. At least now when they interview sucky artists on MTV, they'll have to be honest about where their inspiration is coming from. Instead of "Tom -- the producer -- is such an artist. He, like, totally understood where I coming from on this track" we'll finally hear the truth. "Well, like, you know, Carson, the businessmen at the record company used, like, Microsoft Hit-Maker Pro or something to generate the backing tracks and vocal melody for, like, the entire album, you know. Then, you know, some other businessman brought me some lyrics that, like, you know, test well with 14-year-old girls . .
Of course, the MTV dopes will probably market this as the Next Big Thing -- "Cyber Rock -- the all-digital multimedia experience" or something dumb like that.
take care,
Steve
What I wanna know is, can they come up with a way to make Britney Spears' lip-synching more realistic?
There are really two issues here. One is the simulation of time (rythm), notes, and dynamics (volume). The other is simulation of the expressive sounds of the instruments themselves.
These computer programs all work along the same lines, manipulating rythm, dynamics and notes. I have yet to see any program duplicate the variations in timbre (tone) that a real musician can achieve on his instrument.
I can tell a computer generated score everytime by the fact that the attacks don't vary, and the timbre doesn't change with the volume.
Besides, there's something magical about a group of musicians, all together, when then lock into a groove. It's something that a transcends the abilities of any single human, and certainly any single computer. And that space is where most of the best ideas are born.
On the other hand, working composers banging out MIDI scores on demand will probably find this a very useful time-saving tool. Just lay down kick, snare and hi-hat right on the beats, and turn on the 'humanizer' feature to make it all sound loose. No more twiddling with each note to make it sound right.
Rambling Again,
Loopy
Oh, you mean you have the sound turned on?
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
I'm sorry, no. I really don't think the music scene is equipped for this sort of thing.
Music is interpretation, it isn't an equation. You should not simply use some sort of hokey Scoring + Emotion = Music equation and expect to have something worthwhile come out. Maybe for genuine on-the-fly stuff like context-sensitive game music stuff it could be welcome.
The key here, is art. Music, like other more physical art forms, is all about human imput. If Mike Oldfield's classic 'Tubular Bells' album was released today, and he'd played synthesiser, instead of actually learning and playing that myriad my instruments, it just wouldn't get it.
Even for a group, flexibility is key. I've played with (concert) band conductors who refused to say 'in the concert we will do it this way'. He preferred to deal with the specifics as the situation developed. Things like the crowd are getting bored, let's play this bit faster; the acoustics here really suck, let's make that instrument a bit louder..)
Maybe it would be good for bland elevator/Mc Donalds background music, but not for anything else.
Music is my recreation (listening, playing, creation). Anything that takes that away is bad.
... and today's pet project has
Moving this same type of system to a completely independent composition is obviously an order or two of magnitude more difficult, but essentially, it would require using style fragments instead of music fragments, a way to provide it's own "mood" (instead of being fed by an external source [the game]), and some more complex rules on music theory in general to make things cohesive.
All in all, pretty cool. It'll be unlikely that such a system produces anything particularly innovative or moving, but it will certainly help lower budget TV shows or movies build up better scoring -- it's a sad shame that music is one of the first things cut when budget gets tight. Looking back at B5, Franke's score adds so much to the on-screen scenes, often more than 20 minutes of original music for a 44 minute show...
--
You betray some ignorance of classical music when you say that - not that it's different from the vast majority of people!
Classical music, though usually lacking in the wholehearted improvisation of jazz (though even that almost always relies on pre-existing chords and styles), still involves a lot of variation from performance to performance. The notes might be the same, but that just means that the musicians are concentrating more on the subtle things than on the notes themselves.
As well, just for the record, it's only quite recently that "classical music" (when will we get a better term?!?!) has not been improvised. Mozart, Beethoven, Rachmaninoff, Bach, Paganini, etc.. all improvised - most estimates say they improvised (in concert!) much more music than they ever wrote down. Some classically-trained people still improvise a bit - most decent church organists do, for one. Early music specialists (of which I am one) are always trained in some extemporization.
Perhaps more subtly, but also more importantly IMHO, is that ALL music is improvised when performed half decently - not necessarily everything, but aspects of it.
One example is a saxophone player (mainly classical) who told me in a master class that he takes into account the acoustics of the hall when he's playing. In a boomy, echoey hall, notes that he wants to be short much be played EXTRA short, in order to not be stretched by the reverb. Likewise, in a dry acoustic, notes need to have resonance added to them by the player, as a kind of performing trick, because the room won't create the sound on its own.
Even more importantly than little technical things like this is the fact that any performance is unique, no matter what style. A musician will be playing for the moment - judging his interpretation based on the audience, his feelings, the sound produced, the physical sensations, ambient sound in the hall, etc etc etc... For a truly exceptional performance, all of these things need to be taken into account.
Any music played "without variation" is far from a classical song, in my book! Rather, it's a technical excercise, devoid of music. I'm skeptical of the abilities of this program to do much more than that, though I'll reserve judgement until I can hear it for myself.
It will always take real musicians to do something truely great.
:-)
Just as well as your computer can generate a believable landscape, with natural subtleties, a computer can generate a sound;
Perhaps some day it may progress to the point of resolution where it is very nearly indistinguishable from real musicians.
But, there is always something to be said for those who take art to a new and exciting level.
I am a musician, and I want to hear my computer making music just as much as all you programmers want computers to write your code for you, and be better than you at it.
-stuff to think about
Spoon not. Fork, or fork not. There is no spoon.