torpid plinth
How can I procedurally generate terrain like the one shown in the image? What would be a good approach?
I’ve been trying to recreate this type of terrain using Perlin noise for the past week, but it’s been quite difficult. I came across an interesting article where someone from China recreated a similar terrain in Houdini, but I still don’t understand how it was done.
Article:https://zhuanlan.zhihu.com/p/565348841
I would really appreciate any advice or ideas. Thanks in advance!
crystal aspenBOT
inner drum
Perlin noise (on its own) is probably the wrong tool to use for this.
I don't know how this was done exactly, but if I were going to try and emulate it, I would start by noticing how the scene (mostly) looks like a plane that's been chopped up into cells, and then the corners of the cells have been moved up and down.
Voronoi diagrams are probably where I would start looking for the first step of dividing the cells at all, and after that I would probably be thinking about walking along the edges of those cells, sometimes cutting them loose (to produce the larger cliff faces) and sometimes leaving them connected (where two cells join neatly in a ramp). Sort of like making a bunch of partial cuts in a sheet of fabric, and after that I can start stretching the fabric vertically and see where things fall.
That's just the cliff/terrain bit. Not sure off the top of my head where you would go about generating water/rivers, but there's already enough degrees of freedom here I would want to experiment and tweak things before I knew where I wanted to go next.
No idea if this is actually part of the process used in the image you linked. It could have been done completely differently.
Some loose google translate on the linked article seems to imply a better starting point might be this image, rather than the ones you posted which are significantly further along in the process:
This is clearly a heightmap, so generating this kind of platform-style geometry on a heightmap with some kind of flood-fill algorithm will probably yield closer results. This is probably where the article's rules regarding "wind obstruction" are most directly applied:
Given that this was done in houdini, this might actually be the result of an erosion simulation rather than an additive floodfill process.