On 13 Sep 2005 at 15:53, Chuck Israels wrote: > Also, my impression is that there is an order of magnitude fewer > choices (variables) when one is limited to producing only an > accompanying part. Things seem to get more problematic as more > attention is needed on the "surface' of the music, and programmed, > mechanical, spontaneity falls short of resembling human expression, > at least so far, in my experience.
But I think that's precisely what's wrong with the Wolfram algorithm -- it entirely fails to account for the differences in styles in regard to melody and accompaniment. Because it seems to be generating everything at once, it's more polyphonic in its results than almost all of the styles typically should be. For most styles, it ought to compose a melody line first and then create an accompaniment for it, whereas it sounds to me like in all the styles it's pretty much randomly choosing all the parts simultaneously, with a very small number of parameters specific to each style (certain rhythms and instruments seem to be associated with particular styles). That seems to me to be the reason why the results are so completely underwhelming. Indeed, they all sound rather the same to me, and some of the results from one style sound more like one of the other styles (though that's not always the case). The folks who wrote the algorithm seem to me not to know enough about musical style(s) to be able to design something that will produce reasonable results. They seem to go more for variety and near randomness and ignore harmony almost exclusively, instead of recognizing that music in a particular style is very *conventional* rather than random. -- David W. Fenton http://www.bway.net/~dfenton David Fenton Associates http://www.bway.net/~dfassoc _______________________________________________ Finale mailing list Finale@shsu.edu http://lists.shsu.edu/mailman/listinfo/finale