GMIC Turbulence with the Default Values 32,6,3,0,0, fuzzy, cotton candy sort of pleasingly happy colours

turbulence
turbulence 32,6,3,0,0

G'MIC
TURBULENCE

In theory, a call to turbulence is a call to turbulence 32,6,3,0,0. But a call to turbulence without any trailing parameters only works (for me, because I don't really know what I am doing) if I stop my code short and output to an implicit call to display instead of saving the lot to a file.


For the above image (in PowerShell), I used:
gmic `
  input 1000,250,1,3,0 `
  turbulence 32,6,3,0,0 `
  output image.jpg

turbulence 32,6,3,0,0
turbulence radius, octaves, alpha, difference, mode


I am a fan of G'MIC, but the documentation sucks. I mean, it really sucks. For instance, the last parameter is mode. I assume (as I have not found the place in the documentation where it says) that mode represents the different algorithms which have been implemented for turbulence. And if it is indeed the case that mode is a listing of algorithms (is it), it would make sense to at least (at the very least) give a name to the different algorithms.

Anyway, in the absence of documentation, the best (in my ever so freakin' increasingly humble opinion) way to figure out what the different parameters do is via experimentation.


Radius

Turbulence in this context looks a lot like noise_perlin, patch areas of randomness, which I find soothing, the areas are much small with radius set to one
turbulence 1,6,3,0,0

the splotchy areas are larger with a radius of 100
turbulence 100,6,3,0,0

Unsurprisingly, radius controls the size of the splotches: the greater the radius... er, um, the greater the radius of the splotches.

But if you can tell what octaves controls, you are a better man, woman, child, or nascent AI than I.

Oh, before I move on, I should say that I have a preference (at least in the moment) for radius 100 over radius 1. And as such, that is the setting I shall use for the rest of images on this page.

Octaves

low octaves yields a smooth image
turbulence 100,1,3,0,0

high octaves yields a rougher more nuanced image, with deeper levels of variance
turbulence 100,12,3,0,0

Thus (per the above images), I take octaves to denote the smoothness (or roughness) of the resulting image... the graininess of the turbulence... or perhaps, how many times the function (as determined by the mode) is applied iteratively.

And I was going to say how I liked the smoother image (octave=1) better. But it has less of what I would call turbulence. Certainly, changes to the next parameter seemed to have no effect if the image had already been smoothed out.

Alpha

At alpha we are looking at somthing closer to point noise, like smooth point noise, but the color gradients are smooth, grainy, but smooth, does that make any sense
turbulence 100,12,1,0,0

At Alpha 2, we have the best of the lot, small medium tight clusters with wavy sort of edges, very turbulent looking, sort of like mixing smoke of various colors... or a star nebula
turbulence 100,12,2,0,0

At Alpha 3, we are going closer to the Plasma Gas Cloud, so the splotches are getting bigger, but there is still some of that puffy cloud effect going on
turbulence 100,12,3,0,0

At alpha 4, the edges are smooth, but not too smooth, like polished lacquer on wood, full of color, but the turbulence has all but disappeared
turbulence 100,12,4,0,0

smooth fuzzy edges, like a blur has been applied, pleasing, but not an intricate pattern, one sees it quickly and is done, unlike the Turbulence at Alpha 2
turbulence 100,12,5,0,0

The default is alpha=3, after which point the image becomes smoother and smoother, making it hard to tell the difference between octave=1 and alpha=5.

I really could not tell you what is happening here. But I do know that of the five, I'm liking alpha=2 the best. It looks like some sort of colourized telescopic image, perhaps of a gas cloud.

Difference

This is much like the previous alpha 2 value only the background  is whitish rather than blackish
turbulence 100,12,2,-1,0

I did not note much qualitative difference between difference=-1 and difference=-10. In the trials I ran, negative values produce a lighter (more white) background (like the above) and zero or positive values elicited no change.

Mode

Unfortunately, I could not see much difference between any of the modes... certainly, not enough of a difference to waste the bandwidth posting near identical images. There was a slight difference in lightness. But not much. And if I had to chose a personal preference for any of the modes, I would choose the default of mode=0.

The Recap


Other Functions

I originally started working on this page, as I was doing a write-up on rand, which led me to the various noise calls (noise, noise_hurl, noise_perlin, and noise_poissondisk).

So, any of those might be of use.

Certainly, they are what I am going to explore next.

more
ALPHA CHANNEL
command line image manipulations

© copyright 2020 Brett Paufler
paufler.net@gmail.com