On Fri, Apr 15, 2011 at 12:54:53PM +1000, Peter Chubb wrote: > David> A fixed ratio seems a mistake to me: like optical scaling of > David> fonts, the ratio should depend on the length of the note as > David> well as the speed of the piece. And one does not want to have > David> chords end early in one voice merely because the voice had a > David> longer note starting the chord. > > I agree, but that's really hard to implement. Maybe if Grahame's > virtual violinist can turn into a virtual orchestra, and interpret > music properly (instead of just playing whatever Lilypond produces > like a robotic MIDI player).... but I think that's a long way off.
Yes, it's a long way off. I'm not intending to touch expressive music performance at all in my PhD; rather, I'll make tools which all humans to direct the computer's performance as easily as possible. (which may still not be very easy!) The catch-phrase for this is "let machines do what machines are good at; let humans do what humans are good at". :) Humans aren't good at manually specifying physical parameters every 0.006 seconds, but they _are_ good at correcting the way that notes are played (generally every 0.2 - 2.0 seconds, at least for monophonic violin music). As for other instruments: I'm hoping to get physical measurements of a viola and cello during the summer. It would be a blast to work on other instruments (clarinet, oboe, horn), but that would require physical modelling code for them. At the moment, the only open source code that I'm aware of for clarinet is STK... but as I type this, I recall that although the STK string model uses an algorithm from 1986, the clarinet one is 1998 or 2002. So maybe that's a feasible thing to investigate. Cheers, - Graham _______________________________________________ lilypond-user mailing list lilypond-user@gnu.org http://lists.gnu.org/mailman/listinfo/lilypond-user