0
$\begingroup$

I am trying to scatter some things on a slanted surface and I want to have little to no instances on the slanted part but when I have a normal into a compare node that goes into the selection of the distribute points it makes super sharp edges. Is there any way to make the edges more smooth? And is there a way I could also mix a noise texture into the edge to break it up more and make it look more natural? Current Node Setup

These are not all the nodes but these are the only important ones. I don't want to show them all because this is a crazy complex setup with like five hundred nodes, but these are the only ones that matter, but of course there is a instances on points and a join geometry and the basic stuff.

$\endgroup$

2 Answers 2

1
$\begingroup$

(Using Blender 3.6.12)

This proposal is based on a Boolean Random Value node to draw the probability to instance an object. The probability distribution is controlled by a Float Curve node. The dark grey area of this curve shows the probability as a function of the dot product of the normal with the vertical direction (Z axis). For the demonstration, it is set such that for Value higher than 0.7, the probability is 1. Consequently, an instance is spawn for all faces with the criterion higher than 0.7. Then, the probability decreases sharply with the Value. The mid control point is such that for a dot product of about 0.52, the probability drops to 0.32 (i.e. 32%, i.e. about 1 chance out of 3). It is to notice that in the original GN graph, the Greater Than or Equal node behaves like a Heavyside step function set as Float Curve, with 0 below the threshold and 1 above.

GN Man graph

The following picture illustrates a Shader helping to visualise the criterion (captured as the named attribute "QoI", standing for Quantity of Interest in this demonstration) controlling the probability. Probabilities lower than 50% are rendered in black.

Shader

The Random Value & Mesh to Points nodes combo can be replaced by a Multiply math node computing a Density from the probability to be input in a Distribute Points on Faces node as shown below:

GN with distribute points on faces

Eventually, Random Value and Distribute Points on Faces nodes can also be combined as shown below:

GN with Random value plugged as Selection

Resources:

$\endgroup$
0
$\begingroup$

of course it depends on your geometry on your normal vectors, but you can try it with this node tree:

enter image description here

essentially i "smooth" just the threshold so that it isn't a constant number, but a range of numbers. By this you can get a more random distribution. Don't take my numbers "too serious" - you have to play around with them so that it fits your needs.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .