Thomas de Groot <tho### [at] degrootorg> wrote:
> Op 06/11/2021 om 01:58 schreef Samuel B.:
> > Hi,
> >
> > Here's something I was working on a while back. Zoom to 100% to see all the
> > details.
> >
> Impressive indeed. It /looks/ like a seamless image, is that so? Would
> it be possible to introduce some random size variation in terms of
> bigger chunks of rocks contrasted to smaller ones? The present
> impression is of a rather uniform size, overall, which is telltale for a
> non-natural origin. However, well-done indeed.

Thanks, Thomas.

Good eye. The input height map is indeed seamless, but the entire result is not,
since a 1/3 scaled-down copy was layered over the original.

Yeah, the feature sizes are very regular, unfortunately. This is a problem
caused by the CA I used, but also my choice to scale down the layered copy
instead of scaling up. The CA itself can probably be adjusted to produce larger
features, to a point. But the only real way to make something like this work
would be to make a full-on 3d simulation, not 2.5d like this. But I haven't yet
found a good library to evaluate 3d textures, and the only other way I know
(doing it with traditional arrays [on the CPU instead of GPU]) is just too slow.
I could technically do it, but watching the result in real time would not be
possible since I would also have to perform raymarching using the CPU, thus
bogging everything down.

Sam

Post a reply to this message