MMD > Archives > October 1995 > 1995.10.26 > 03Prev  Next


Re: Genetics (Analog to Midi) and Re: Music Box Designs and Re: Me and the Convention
By Larry Smith

This was originally send Oct 13th, but it bounced the week
I was gone.  Here it is again.  I will be following up
on this once I have caught up my mail.

 - - - - - - - - - - - - - - - - -

To: rolls@foxtail.com (Automatic Music Mailing List)¶
Date: Fri, 13 Oct 95 14:15:32 -0400¶
From: larrys@zk3.dec.com

Subject: Re: Genetics (Analog to Midi)

Karl_Petersen@smtp.scpboi.com:

re: genetics

>     The next question is, what hardware and software are necessary to give
>     this a trial run? I have lots of overnight time available on DOS and
>     UNIX workstations at home, and have my own resident AIX administrator
>     and C-writer, but he has not osmosed the genetic algorithm recommended
>     reading list.
>
>     Ok, Larry, Tag, you're It =).

The most important software task is to design the comparator, because
that is what will define how well the algorithm will work.  That will
probably merit a discussion about what and how to measure success.

Beyond that, you need a way to suck up a target performance as a .wav
file, and a way to create a .wav file from a midi performance short of
just playing the thing out loud and digitizing the result - that would
work, but it would be mighty annoying to listen to while it worked, I'm
here to tell you.  Other than that, you need just a basic genetic algor-
ithm, a number of them have been published in various computer rags and
I expect a short conversation with a gopher, or perhaps a web search,
will quickly turn up some archives with C++ code implementing a reason-
able algorithm.  Almost any hardware will suffice, but faster is obviously
preferred.  If you need a sound card to do these conversions a Pentium
would be a logical choice.

The comparator will also implement your contraints - restricting target
voices, for instance.  The algorithm will be faster and more accurate
if the output voices are similar to the input voices - which is why I
came up with the idea, I should be able to use it to take recordings of
musical boxes and turn them into midi fairly easily.  But taking something
like, say, "In a Gadda Da Vida" with output set to a music box voice may
lead to something a bit...er...funky.  Some other constaints we'd like
would control the number and timing of notes - the algorithm will call
for 86 notes at prime-number picosecond intervals if that will better
mimic the target .wav - and while that may be fine with the midi-to-.wav
simulator, it would be hellish to try to pin, dontcha know.  To do a
straight whatever-to-midi will be easier and harder - easier because
the voices will match, harder because the algorithm won't care if some-
thing is hard to play provided it sounds right.  Some "post output"
scanning might be called for that would do things like scanning a bar
of 60% violin notes and replacing the odd cello note with the equivalent
violin notes.

Subject: Music box designs

I was looking at tremolo boxes the other day - these are (I think) the
ones with duplicate notes in the comb in order to achieve a sort of
sustain.  I was wondering if anyone had thought of using spinning
disks to "bow" comb teeth to accomplish this, it would seem a fairly
logical progression from classical music boxes and might make for a
really lovely sound, sort of like a cross between a reed organ and a
music box.

Subject: Me and the Convention

Well, I'm off to soggy Florida for a week at Walt Disney World.  I might
still drop by the public mart at the convention even though Nancy Latti
cleaned me out of my total acquisition budget for the next two years,
so I might see some of you folk.  I should have plenty of mailspace on
this node, so keep the digests coming, and I'll see you folks Monday,
the 23rd.  Looking forward to hearing everyone's take on the convention.
If anyone here goes to Charles Hind's workshop on building new musical
boxes, _please_ collect whatever material you can and post a full review
of what he's been working on!

regards,¶
Larry

 [ Larry,
 [ Its too bad this last part didn't get here in time.  Did you talk
 [ to any of our subscribers while you were at the convetion ?
 [ Jody

(Message sent Mon, 23 Oct 95 09:36:06 -0400 , from time zone -0400.)

Key Words in Subject:  Analog, Box, Convention, Designs, Genetics, Me, Midi, Music