Let say if I want a sphere to have some natural shape, I fetch position data of point or voxel to turbulence vop and add that to my global position ,right?

I saw 1 example and he get the vectors which is shooting from center of the sphere to the current positions (basically shoot from inside to outside) and normalized the vector and multiplied to the turbulence value before adding the value to the global position.

I've heard that is to make give turbulence value only for the edges of the sphere but how that actually work out?

Hi,

Think of the point position as a vector that describes a displacement (from the center of a sphere to the point's position on the surface of the sphere). When you multiply a scalar by a vector you're affecting the magnitude of that vector, but not the direction. In the case of a point on the surface of a sphere, you'd be changing it's distance to the center (radius) but not it's angles. A noise VOP will return a “random” value based on the value you feed in, so for each point you'd be getting a different output from the noise VOP, which multiplying the point's position would return vectors with different magnitudes (hence different radius). In the case of a sphere you're basically applying a displacement from the center of the sphere.
For volumes it's quite different though, because voxel values are all scalars, so by multiplying a noisy volume by the volume of, say, a sphere, you'd be eating out density wherever the noisy volume is 0.
Does it make sense?

Cheers

Thanks for the well explanations.

I understand the directionality of turbulence for the points.

Howeber, for volume, what would be main benefit with multiply normalized position data of voxel to turbulence and add to position to get distance and remap(fit) for getting edge smoother
and
just multiply turbulence with position data of voxel and get distance and remap(fit) for getting edge smoother?

I don't see much difference.

fkkcloud
Thanks for the well explanations.

I understand the directionality of turbulence for the points.

Howeber, for volume, what would be main benefit with multiply normalized position data of voxel to turbulence and add to position to get distance and remap(fit) for getting edge smoother
and
just multiply turbulence with position data of voxel and get distance and remap(fit) for getting edge smoother?

I don't see much difference.

Yeah, I suppose you can indeed advect density like that. I'm under the impression it's not always predictable, but I'm probably wrong :-)
You might want to check SDFs though…

Cheers