i am working on a project which seeks to record
say a conversation and transcribe it into a score. basically
find a way to find the notational equivalents to the sounds
so that they can be mapped out into into time signatures
etc....

any ideas by which to approach this...?

i know that there is software out there which takes midi
and transcribes it...so i first thought i could begin by
trying to convert the audio into midi and do it that way.
and i know 'fiddle' does this, just never really used pd
and wanted to see if anyone could help find suggest the best
way...

any help much appreciated

travis

[fiddle~] is actually a good choice I think. You could split your conversation into two streams, lowpass one below 1kHz and use it with [fiddle~] to get a pitch contour. Severely highpass and gate/expand the
other to get the stops and fricatives as event markers. Not sure how you would best combine these
to get a score
You might like to ask Mr Hovagimyan how he did this in Max
http://homestudio.thing.net/forum/arch/msg00355.html

you might also check out scrabledhackz (http://www.popmodernism.org/scrambledhackz/) the developer hasn't released the code yet, but it looks to be a comprehensive sound databse compiler/sorter. For taking random sound input and converting it to discrete events which can be scored, I think this scrambledhackz will be very key.

there is also CLAM (http://clam.iua.upf.edu/) which includes the annotator. It does complete analysis of complex chord structures from recorded sound. It's off the a)hook cuff c)grid.

Sunji, CLAM looks totally awesome, thanks for the heads-up!