> Hi.
>
> I have a bare-bone melody dictation utility that generates "mudela"
> output (Linux). At the moment it's just a demo with little practical
> usefulness: you start it, it listens to the mic for a selectable amount
> of time, then prints out the pitches you played or sang, with no
> duration.
Interesting! At our Signal Processing group at the Royal Institute of
Technology, Stockholm, we give a project course where the students
solve different problems using a DSP. This year, we actually had a
group that implemented a similar program, though not connected to
Lilypond. Unfortunately, their solution wasn't that robust, I hope
yours works better.
> Internally the programs has an estimation of the note duration so that
> it is possible to generate a complete transcription of the melody. I was
> wondering what is the best way to do this.
> I see two possibilities:
>
> 1) The user declare the duration of the first note, then the subsequent
> duration are quantized and scaled accordingly.
>
> 2) The user declares the tempo (like in a MIDI file), then he sings (or
> plays) "a tempo".
I haven't tried any of the commercial music typesetting programs
or sequencers where you can use a synth keyboard to enter the music, but
as far as I understand, most of them use the second alternative.
However, I've been thinking of a solution like 1) where you also
adaptively try to track small variations in the input tempo. Of course
you have to use user parameters that tells what note durations should
be considered, just as in midi2ly.
(Maybe you should consider to rename the program in spirit of
the mi2mu -> midi2ly change that was done recently.)
> Does anybody know how to embed such a program in an editor? (Emacs?)
Shouldn't be too hard to make a lisp function that calls your
external program and includes the output in the buffer.
/Mats
