Re: Denemo-feedback

[Richard Shann]

On Wed, 2013-12-04 at 23:37 +0100, Johan Vromans wrote:
> Richard Shann <address@hidden> writes:
> 
> > At the moment this AI step is the simplest possible, it goes through the
> > durations that you played assigning them to the nearest note lengths
> > using only whole-note ....1/256th note and dotted versions of them (no
> > triplet values have been entered in the table of durations).
> 
> I assume this is customizable to, e.g., whole-note...1/8th or something
> similar. 

I guess you mean the list of durations and the algorithm to choose
between them? That is still in C but it is pretty trivial to export it
so the user can play around with the AI bit.

> 
> But even if only 80% of the durations are correct, this would already be
> helpful (for me, in some situations).
> 
> I'm not a piano player. I can, however, enter notes using a (MIDI)
> keyboard in a slow and rhythmless pace, and I can tap the correct rhythm
> on a single key. I'd love to combine these two inputs.

Well, I think it would be a good idea to test this out before I start
writing code to support it. You can do this by writing a melody that is
all on one note. That is, enter a piece just by tapping on one note.
Then look to see how much of the durations it got right, and see if
accepting the correct ones and altering the wrong ones would be a
practical entry method, ignoring the fact that all the notes are the
same pitch. (You can, in any case add these over the top afterwards just
by placing the cursor on the first note and playing the notes on the
MIDI keyboard at your own pace).
I'm sure the first thing you will want is to start playing with the AI -
are you able to write simple Scheme? If so I will break this out for you
to play with sooner rather than later.

Richard