Re: MIDI to sheet music

[François Pinard]

[Han-Wen Nienhuys]

> address@hidden writes:
> > (Question for gnu-music-discuss: Does anyone know of a way to improve
> > midi2ly's performance, or is it just something that needs work before
> > it can be a really useful tool?)

I only use it two or three times, and felt the need of doing some
post-processing of the output score, in Python.  I remember having thought
that these things could probably be done right in the program.

> The rhythmic quantization of midi2ly is quite naive; you might want to
> look into signal processing techniques (wavelets, windowed fourier
> transforms, etc.) to do quantization more precisely.

Agreed that it can get rather complex.  Surely an interesting project to
submit to one of your friendly teachers, in exchange of getting rid of
other year's duties :-).

Automating the proper choice between flats and sharps also looks like
another difficult task.  You not only ought to cleverly guess the composer
intent about tonalities and modulations, you also ought to know the rules
(or usage!) pretty well in things like chromatic transitions.

> On the whole, I'd appreciate a rewrite of the program, perhaps in
> Python. The C++ approach is rather clumsy for what it is doing.

It would surely be an interesting research exercise, to say the least, and
Python might be proper the proper tool.  Ah, if I only had many lives ahead!

-- 
François Pinard   http://www.iro.umontreal.ca/~pinard