RE: [Rosegarden-devel] Re: [Rosegarden-user] plans for notation

[Mark K. <mk...@co...>]

> -----Original Message-----
> From: ros...@li...
> [mailto:ros...@li...]On Behalf Of Chris
> Cannam
> Sent: Friday, January 03, 2003 2:11 AM
> To: Mark Knecht
> Cc: Rosegarden-Devel
> Subject: Re: [Rosegarden-devel] Re: [Rosegarden-user] plans for notation
>
>
> On Friday 03 January 2003 05:53, Mark Knecht wrote:
> > a sax line this evening in notation. It was just sort of a run
> > through the scale, but played sort of staccato. What I was getting
> > from RG was very short notes and very short rests, as opposed to
> > maybe 16th notes with a staccato indication.
>
> There's fundamentally no way to tell the difference in MIDI between a
> staccato 16th note and a 32nd-with-a-32nd-rest.

So possibly a configuration option that tells you I'd prefer a staccato
representation, while Guillaume prefers rests? Presumably RG will eventually
be happy doing both. If they are truly interchangeable (I'm not sure they
are) then you'd have to ask which is preferred, I think.

The other question is whether RG makes and use of MIDI velocity today? Is it
possible that velocity information, either overall or within a group of
notes, would help?

I don't know that you'll be up for this next one, because it's a huge issue,
but for the sake of talking here goes...

Understanding that humans are by our very nature unpredictable, then we know
that even the very same phrase on paper played 8 times is going to be unique
every time in the MIDI data. With this in mind, I think it's important to
have a long term strategy that takes this into account. Presuming that you
agree, at least for the sake of conversation, then I'd suggest that RG needs
to be able to look at the whole score to make a decision how to draw even
one part correctly. This leads me to see at least a two pass optimization

I'd imagine that in real time, on a first pass, the score might look very
much like it does today. However, if the program was smart enough to
recognize patterns then it could gain a lot of information by averaging
multiple instances of the same phrase. I'm seeing this sort of problem in my
sax line right now. There's a little triplet or something that is getting
notated differently just about every time it comes up, even though I intend
it to be the same on paper. They all get played a little bit differently, so
RG does the best it can within a limited distance. If RG looked at 20 of
these phrases as a group, on a second pass optimization, then I suspect that
the quality would probably go up dramatically.

Getting even a bit smarter, I can imagine that RG can see phrases based on
how the note relationships are to each other, as opposed to just the notes
themselves. Key changes, modulations. The tune that comes to mind is the
very opening of 'A Little Night Music'. The hook (yes, Mozart really did
write hooks, didn't he...) is repeated twice, but starting at a different
place the second time. If RG could 'see' that it would be much more likely
to extract something meaningful.

A tremendously interesting problem.

>