The first stage, generating the coordinate texture,
has been simplified by using geometry nodes
which are intended for manipulating normals. 4)
The new setup has fewer nodes, supports rotation,
and has explicit inputs for center, angle, and size.
The first version required planning and math
to make renders of different parts of the set.
The explicit inputs in the new version
make it easy to predict and control
the new image that will be rendered.
Self-similar images, like the Mandelbrot set,
look very good in looping zoom animations.
The new coordinate inputs make these easy.
Keyframe a starting location, size, and rotation,
keyframe another for the end, and render.