Re: [OT] Vivi, the Virtual Violinist, plays LilyPond music

[Graham Percival]

On Tue, Mar 08, 2011 at 11:53:16AM +1100, Peter Chubb wrote:
> Hi Graham,
>    Vivi is WAY COOL.  I really like the way you
>    start with a physical model of the instrument.

Thanks!  Yes, although software like Vocaloid is neat (and the
user community is incredible!), I always feel a bit "let down" by
sampling-based approaches.  Physical modelling is just so elegant!

>    I was also interested in the training aspect, as it links in
>    with Deborah de Graaff's work here on how eminent musicians
>    practice and learn new pieces. See
>    
> http://marcs.uws.edu.au/links/ICoMusic/Full_Paper_PDF/deGraaff_Schubert.pdf  
Interesting, I hadn't come across that one before!

>    And I'm sure you've also seen the automatic music performance work
>    from KTH?  http://www.speech.kth.se/music/performance/download/ 
>    It's all open source.

Director Musices?  Yes indeed.  I've been musing on integrating
that work.

So far, Vivi doesn't do any "expressive music peformance".  I'm
aware that it's an active research area, especially with Rencon:
http://renconmusic.org/
(which is taking place at the same conference I'll be presenting
Vivi at)  but I haven't done any work in that field yet.  Right
now, I'm trying to bring automatic violin playing to the level at
which the expressive music performance work can be applied to the
violin.

For the past few years, almost all work on expressive music
performance is done with piano, with a little bit on guitars.
Both of those are "fire and forget" instruments -- once you hit a
piano key, you don't need to change any parameters.  That makes
them a natural target for algorithms focusing on note-level
musical expression... but as a string player, I wish that more
work was done on violin and cello music!  :)    I'm hoping to
automate the physical actions of violin playing (i.e. the things a
player does at around 100 - 1000 Hz), so that other researchers
can focus on the note (and phrase)-level musical expression (i.e.
decisions made at the 5 Hz - less than 1 Hz) level.

Of course, those "other researchers" could well be me in 6-12
months!  It all depends on how quickly I can get Vivi playing
chords and double stops, doing vibrato, etc etc.

Cheers,
- Graham