One method is to pseudorandomly generate points on the 2-sphere, and then repulse and normalize the points.
https://mathworld.wolfram.com/SpherePointPicking.html
Once the points are generated, something like qhull can be used to generate a convex hull.
taby said:
One method is to pseudorandomly generate points on the 2-sphere, and then repulse and normalize the points.
Nice to eliminate any mathematically constructed appearance, and looking more natural.
But we get high variance in triangle areas, and the repulsion to fix this needs many iterations.
An alternative is to construct poisson samples over the sphere. I did this to create simple procedural rocks, with a low number of faces like 16:
Just - i generate faces per sample, not triangles. Imagine the center of the faces and connect them to triangles. The result looks similar to icosahedron, but there is no perfect and constructed symmetry.
Useful if we want natural stuff.
But in both cases we also need a Delauney Triangluation over the points to find the proper triangle tessellation, so all this is much more work.
