1

i am working on a project in blender 3.12 i have instances scattered on an object and i want to "blur" the normals of these instances. i need variable blur, so that in some areas the instances have the unchanged normals and in other areas nearby instances normals are more averaged. unfortunatley attribute blur or sample nearest do not exist yet. can anyone think of a cheat to do this? it doesn't have to be properly blurred, just sorta averaging the nearest instances (with attribute statistic node, mean), with the "blur factor" affecting how many nearby instances are take into account for example

Walter
  • 329
  • 2
  • 10

1 Answers1

2

Assuming you mean version 3.1.2, as 3.12 version doesn't exist…

What is the difference between Blender Version X.Y.0 and X.Y.1?

…The possibility to sample nearest already exists. We used Transfer Attribute Node back in the day - those of us veterans, that are still alive anyway:

Consider studying this thread to see what is available where:

Can't find the node! Which node is available in which Blender version?

Also why are you stuck on 3.1.2? Can't you open the project in a newer version of Blender, it stops working for some reason? As argued in the first link, you should be able to upgrade…

Alright, so here's how I would blur stuff back then:

I'm aware that the blurring would be better if I switched the group output domain to points, however I started with faces to get a sharp result first, and changing the domain to compare results would be cheating, the purpose of the GIF was to show the blurring of just interpolating between points and faces again. I don't know if it works in your case because you didn't present one…

As for controlling the blur, you can position Mix nodes in various places (not just at the end as I show) to control that:

Markus von Broady
  • 36,563
  • 3
  • 30
  • 99
  • If you don't have a starting geometry you can create one, though it might be tricky in 3.1.2. for 2D geometry you can spawn 0-length curves on points and then fill the curve. Sometimes convex hull works. You can also use old algorithms for sorting to get $n$ nearest points for each point ($n$ equal at least 2, maybe more for more relations). – Markus von Broady Sep 02 '23 at 18:18