Generating surface features from OpenSimplexNoise

MirceaKitsune

I'm unsure how to approach the next step in my marching cubes terrain generator. I have each chunk reading the appropriate noise values in its cubic area, fetched from a dictionary of points of the form points[Vector3(16, 32, 64)]: With everything else pretty much optimized the result is predictable and looking good!

As you can see I get an infinite maze of caves. It's pretty but I obviously want this to only happen underground: At the surface, starting around height level 0 (water level later), I wish to have a flat surface with casual cave entrances otherwise flat meadows or mountains... not counting oceans yet as water isn't implemented but the biome for it should ideally be there.

Obviously this happens because I use read from one noise object directly. My question is how should I interpret this noise, possibly mixing it with another noise, to get a proper landscape at the surface? In other words what should I add to:

var noise = OpenSimplexNoise.new()
noise.seed = randi()
noise.octaves = 3
noise.lacunarity = 2
noise.period = 16
noise.persistence = 0.5
func _get_noise(pos):
	return noise.get_noise_3dv(pos)

pkowal1982

Try

func _get_noise(pos):
   return noise.get_noise_3dv(pos) if pos.y < 0.0 else c

where c is value between [-1, 1]

MirceaKitsune

I found a solution, seems to be working better than I expected for being as simple:

@export var density_curve = Vector2(16, 1024)

func _get_points_noise(pos: Vector3):
	var ofs = density_curve[0] if pos.y >= 0 else density_curve[1]
	var n = noise.get_noise_3dv(pos) + (pos.y / ofs)
	return n

This lets me customize the density offset above or beyond the 0 point independently. In this case I can go down roughly 1024 units before caves are too dense to keep exploring, whereas plateaus above ground level have a maximum height of 16.

This doesn't give me proper hills and mountains of varying sizes like I want, but for now they provide a good enough result for testing. If anyone knows how it can be improved to make the result more complex while keeping the code simple do let me know.

cybereality

Love the aesthetic!

MirceaKitsune

@cybereality said: Love the aesthetic!

Thanks! I'm impressed how far I managed to get with it overall. I need to do a few more cleanups then I think I should have something worthy of putting up on Github.