>I am beginning with timing of the notes, ignoring the tone for the present.
>It's a simple matter to design a program to create random notes of proper
>duration fill the measures. The challenge is to create an artificial
>intelligence that makes the sequence of notes "musical." I suppose one
>could compose atonal music, but that's not my thing.
>
>Any help or advice will be appeciated.
>
>TIA
>
>Richard

Richard,

Cool project. Being a graphics guy typically, I would suggest using a
pixmap as one method of early note generation. I have a posterization
filter I can send that you could be recoded to generate discrete levels
which are then used to lookup note data. There is some amazing software out
there that uses the pixmap method already. You can use brightness for
instance (imagine a gradient) as a method for velocity (volume) or pitch as
well, a portamento.

I would start otherwise first with octaves, just bounce between them, then
add 5ths,7ths,3rds.

You could also have a lookup table for chords,etc.

What I would do were I writing my own program like this is to first teach
the program what sounds "bad". These combinations could then be stored as a
lookup table for remapping if encountered or perhaps sent back as one level
"better" as "bad" can have some levels in itself.  A  "C" and a "D" played
at the same time can sound dissonant for instance when side by side, but
will often sound very fine when used in a chord an octave apart.

Hope this is of some inspiration to you. Sounds like a great project.

Eventually you will want to add a little random swing factor for the timing
too to make it more human sounding.
For making "pop" formula music, there is the standard formula, intro-
verse-chorus-verse-chorus-bridge-verse-chorus-end. Or bridge, chorus-end.

Music is math done right in a way, so I would study a little of the math
contained in something that sounds good to you.

Robert Covington