What do Presets react to?

[Gerdski]

Hi all,

I am checking out Presets now to find out which ones are best suited for (my) live music. It seems to me that most of them react to changes in volume or transients, this is probably what "beat detection" means. Some of these appear to be linear, so that the change in visuals increases with the difference in volume.

Are there any presets that react to pitch or notes?
Or maybe the panorama position of a stereo signal?
Can I somehow see what my favourite plugins react to?

Thanks!

[Nitorami]

Virtually all presets react to changes of bass, mid and treb signal. These signals are what their name implies, and sum over left and right channel. A number of different schemes have been devised, using different weighting, time constants and other mysterious factors but basically they all react to changes in these signals. Some try to create a trigger signal (a "beat") on which they react, others change zoom or other movement or distortion parameters by volume (or by any mixture of bass,mid,treb; by a weighted mixture; by a nonlinearly weighted mixture, using sqare or square root or whatever else).

A qualitatively different scheme is my own harmonic beat detection. It is still based on bass,mid,treb, but rather than reacting instantaneously it tries to fall into pace with the music and find the rhythm, like a pendulum. It can be quite impressive but only works well with rhythmic music, may take a few seconds to find the rhythm, and it may appear off-beat, therefore I used it sparingly.

AFAIK there are no presets that react to pitch or notes (other than showing a waveform or harmonic spectrum). Left and right channel have neither been used either, and none of this would have been possible anyway with milkdrop before Benski introduced the ns-eel language extension in 2009.

You may of course study the code in the per frame section but trust me that will hardly give you any insight. The appearance of presets depends on much more than the beat detection, and some of the schemes devised are so convoluted that you have no chance to understand them anyway.
Give me a few examples of what presets suits you, maybe I can tell you more.

[Flexi]

I have done some experiments with spectral volume binning from more bands than just the usual bass mid and treb values.

You can, with certain effort, pass spectrum curve values from custom waves over the gmegabuf data structure to the rest of the Milkdrop scripts.

Check out my preset "target practice"

http://forums.winamp.com/showpost.php?p=2730448&postcount=190

and with fancier graphics, "alien fish pond":

http://forums.winamp.com/showpost.php?p=2666985&postcount=181

I'm sampling from custom wave 2. In its per-point code, you can find the filling of gmegabuf with the volume values from the 64 samples.

I organized the sequentially indexed gmegabuf with multiple instances of a number of kinematic agents with a list of properties (and some additional space to build up an index of their k nearest neighbors too). See in the initialization code block the count variable for the actual number of objects. beware the rather complex KNN algorithm implementation in per-frame .

The agents get accelerated each by their own band's volume. The bigger agents sample from the lower frequency spectrum bands and the smaller ones from the higher pitches.

[Zylot]

I'm pretty sure, just as an aside, that value1 and value2 in custom waves have been the left and right channel since they were introduced.

[ryan]

I find it nuts how far Milkdrop has come since I was 14