rhodes@foxtail.com (Robbie Rhodes):

> Howzat again?  Yes: true random noise on the time base doesn't sound right,
> either.  Therefore, the original performer was NOT random in his timing,
> and there must be some sort of correlation with the musical figure he's
>performing.

Yeah, that doesn't surprise me one bit.  I'd expect this noise to
come from two sources, one being "expression" expressed with minute
changes in tempo based on the material being played at the moment
(and affected by the mood of the musician, background noise, and
so forth and also affected by background time marks, such as clocks
or clapping of the audience - even cadences of steps of people that
are moving around).

The other source would be smaller, but would relate to the player's
physical ability to move to the next note - i.e. whatever the metro-
nome marks, it takes physical time to get ready for the next note,
which may occur slightly ahead or slightly behind depending on how
quick it is to reach.

I suspect "live" performances sound that way mostly because of the
first effect, there is a real interaction between player and audience
that cannot be replicated.  However, one of the real charms of mech-
anical music is listening to the "enthusiasm" of the player despite
what (if any) enthusiasm is coming from the audience.  I'm thinking
of a "can-can" piece on a Welty I heard on CD once, whose undeniable
enthusiasm still makes me smile when I think of it.  This is mostly
volume and fast tempo, I suspect, so using a faster arrangement and
more velocity will help.  The second effect, minor timing variations
due to physical instrument manipulation, _could_ be simulated if you
did an analysis of access times for each note and the effect on timing
and velocity.  Not a trivial exercise by any means.  _IF_ we managed
to create a genetic analog-to-midi translator, however, it would not
be unreasonable to post-process the midi output and analyze the
timing variations to get a better idea of how to synthesize these
effects using a relatively simple mathematical algorithm.  One of
the great strengths of genetic programming is that you don't need to
fully understand something in order to program with it.

Good stuff to think about, though.  Just a few ideas off the top
of my head.  Should get some good feedback from the list, but I
would also suggest posting it to comp.music and archiving the
thread so Jody can put that on the list as well.  I suspect there
are some people who could really run with this on that newsgroup.

> P.S.  I love your enthusiasm on the music box stuff.  Keep talking,
> keep writing, keep your spirits up.  I'll write a note soon suggesting
> a method of Midi control for your traditional mechanism.  (Hint: its based
> on another widespread multi-channel control mechanism that's almost as old
> as the pinned cylinder!)

Enthusiasm is one of those things I bring to any endeavour that
interests me.  =)  But, as I noted above, there is something about
mechanical music that I just find much more engaging than a modern
CD of some performance, no matter what one says about the quality
of the reproduction.  A mechanical instrument _is_ a _real_ instrum-
ent, and that makes all the difference.  Looking forward to hearing
your next idea...

regards,¶
Larry