[linux-audio-user] Progressive Quantisation (long)

[Larry Troxler]

On Tuesday 13 June 2006 18:20, Mark Knecht wrote:
> On 6/13/06, Folderol <folderol at ukfsn.org> wrote:
> > Well I don't know if this term actually exists or if I've just invented
> > it!
> >
> > This is an idea I've thought about for quite some time, years in fact,
> > but don't have the programming ability to try to put it into practice.
> > As I think it should really be part of, or a plugin to a sequencer I've
> > posted to LAU & Rosegarden lists. I hope nobody minds. I'd be very
> > interested in other people's thoughts on it.
>
> <SNIP>
>
> > What I would like to see is quantisation algorythm the detects trends
> > rather than absolute values, then progressively applies small
> > corrections the keep overall timing correct. (it would of course have
> > to operate over all tracks simultaneously).
>

So you have this tempo-imprecise piece of playing that you've recorded, and 
now you want to quantize it in such a way that the resulting performance 
follows a strict metronome throughout the whole piece?

If so, then might I offer a thought here? it seems to me that what you are 
asking for is a beat detector, followed by tempo editing.

I think that this would be a more general purpose feature set that would still 
accomplish what you want to do. Consider that often (at least in my case) I 
would like to detect the beats, but _not_ to correct the tempo, since to me 
at least, the slight variation in tempo sounds natural and captures the 
original spirit of the recording. However, I might  like to add additional 
tracks that are synchronized to this meandering tempo, and also would like a 
notation display of the tempo-meandering performance.

So in my case, I would run the beat detector, which would generate a tempo map 
and adjust the event times> The end result would sound the same as the 
original performance. In your case, you would take the additional step of 
then editing the tempo map to clear out all the generated tempo changes. That 
is, if I understand what you want to do correctly.

Of course, there's always the possibility to provide an option in the 
sequencer's beat detector dialog, to not change the tempo map, and only 
adjust the event times.

There is some prior art in studying beat detection algorithms - consider 
http://www.ofai.at/~simon.dixon/beatroot/index.html

> Anyway, the thought I was having had to do with the need to discover
> which hits were intended to be on the beat, whatever that is, and
> which hits where these ghost notes. IMHO a good tool for changing MIDI
> tempo would determine which is which and would vary the tempo without
> making huge changes to the time between the main hit and it's ghost.
>

Would this in practice be an issue if the tempo doesn't vary wildly? If it is 
an issue, I guest there are several possible approaches to deal with it. And 
exercise for the reader, as they say? :-)

Regards

Larry