Square Voronoi?

Started by René, July 18, 2019, 12:15:55 PM

Previous topic - Next topic

Dune

It can certainly look good (as you show again), but if these noises take so long to compute, it's off-putting too. I think once you have a hang of what works in terms of greyscale and blurryness, it's quite easy to make some standard maps to use.
These sites offer some quite interesting shape generators.
I hope I will be proven wrong though  ;)

WAS

Quote from: Dune on August 01, 2019, 08:42:05 AM
It can certainly look good (as you show again), but if these noises take so long to compute, it's off-putting too. I think once you have a hang of what works in terms of greyscale and blurryness, it's quite easy to make some standard maps to use.
These sites offer some quite interesting shape generators.
I hope I will be proven wrong though  ;)

Euclidean is just as slow as Manhattan from all the tests I did in other languages, and it's noted in documentation about the raw formulas and machine iterations, that they are both slow. So it should be nearly if not just as fast as Voronoi Diff Scalars when incorporated correctly.

Seems the slowness in Hetzen's function is the blue nodes speed themselves when being calculated.

WAS

#77
Quote from: mhaze on August 01, 2019, 04:54:29 AM
I think I prefer a squared worley noise to Manhattan - See illustration. OR a mix of both!
and
https://thebookofshaders.com

Those are just really circles though. Use Perlin Billows or Voronoi Billows or inverted Ridges. In any case, we've had these a long time and I use them all the time in displacement, which do not create squares.

Their terrain example they haven't gotten into has a good example I'd like to use: https://thebookofshaders.com/edit.php?log=161127200501 (play with the first fraction function for a seed)

The canvas scales to viewport, so you can go into a high resolution and maximize the canvas and save the image.

René

It remains to be seen how fast or slow Manhattan is, and even if it is slow, everyone is free to use it or not. As said: the Alpine shader was also slow in the beginning, there was even a warning included because of that. And computers are getting faster. :)

WAS

Quote from: René on August 01, 2019, 12:46:09 PM
It remains to be seen how fast or slow Manhattan is, and even if it is slow, everyone is free to use it or not. As said: the Alpine shader was also slow in the beginning, there was even a warning included because of that. And computers are getting faster. :)

Hetzen's latest function is effectively a working manhattan noise function. It is slow. We have yet to see what Matt incorporates, yes.

WAS

@René have you actually tried mixing together Hetzen's cobble on Y instead of Z (for lateral)? I imagine with some mixing between flatness of cobbles, and some better warping than this quick test with soft perlin warps you could come up with some decent square rocky texture to apply to shapes.

Warping in my quick example is a little too soft imo, and I feel there could be another mix of cobble into there, as well as masking some of the cobble out entirely to create more variance so it's not so square-y.

René

Oh, yeah, I tried. ;) Sometimes with promising results. The biggest problem, however, was to get more (natural) variation. This was partly because I don't understand much about the node setup and partly because the regular pattern kept returning, no matter how many gobble stones I mixed.
I also realized that having square noise/voronoi is only half the solution to obtaining credible block shapes, although it is certainly easier to do.

René

I'm also curious about Manhattan noise's ability to make terrains. A blocky terrain is essential to make credible square noise.

WAS

Quote from: René on August 22, 2019, 02:51:43 AMI'm also curious about Manhattan noise's ability to make terrains. A blocky terrain is essential to make credible square noise.

Same. I hope it's added.

Your Mesa Squares looks very nice.

I tried some different methods to mask out the squares, so they weren't complete, but it actually proved harder than I thought. Masking them out-right leaves you with holes in the displacement lift which is hard to deal with when mixing different levels of the pattern. I than tried overlaying the same level of scalar into the masked out areas to try and hide cut outs, but that wasn't working right with the later mixes. Difficult stuff, but can come up with some interesting shapes. I think it would honestly be good for those broken up blocking sedimentary layers between larger Manhattan-type layers.

WAS

Hey @Matt  is there any update on a native Manhattan noise? I wonder if this topic is of any help with ideas: http://www.pixelninja.design/manhattan-voronoi-approximation/

Matt

I don't have an updated on that at the moment.
Just because milk is white doesn't mean that clouds are made of milk.