Yann Le GallMay 27
playing with this concept over the weekend. I like how the colors/glow turned out.
Show threadLeonard RitterMay 28
@ylegall getting isolines as curves to move oriented quads along is next level stuff. i wouldn't even know where to start.
Show threadYann Le GallMay 29

@lritter Thanks, the software I'm using (Houdini) has some nice utilities.

For example, Houdini provides a "copy to points" operation which can transform arbitrary geometry onto points with attributes to control orientation, scale, etc. so it is essentially building a 4x4 matrix to transform vertices.

Show threadLeonard RitterMay 29
@ylegall thank you. what you do once you have the points was clear to me but how do you get the points?
Show threadaevaMay 29
@lritter @ylegall I assume it's like blender's "distribute points on faces". you just delete the ones that aren't near the isolines and the ones that are too close to each other?
Show threadLeonard RitterMay 29
@aeva @ylegall you can snap them with a distance field but how do they move? you would have to take the gradient and pick a tangent which is easy on a 2-manifold; yet you have two possible directions: cw/ccw
Show threadjungkopfMay 29
@lritter @aeva @ylegall
In a previous comment, the algorithm for building the isolines is described. I‘m guessing that the points of a line are then converted to curve primitives and evenly resampled using Houdinis „resample“ node. Houdini easily allows you to assign a value „u“ along a curve ranging from 0 (start) to 1 (end). This value can be used to look up the global xyz-position for a point at „u“ along the curve. Incrementing „u“ for every point moves it along the curve.
Show threadmmby
@jungkopf @lritter @aeva @ylegall yeah my guess too - the isolines don't change, so the isolines could be approximated by b-splines on the manifold, then animating is easy (since you know the arclength for the animation, which would be hard/impossible(?) to parametrically describe)
May 29 at 8:02amWeb
Show threadaevaMay 29
@mmby @jungkopf @lritter @ylegall yeah, the sphere is just rotating. convert points with similar sdf values to splines, distribute points on curves to get the nice uniform distribution, animate them along the curve and rotate the sphere to make it look like its morphing
Show threadLeonard RitterMay 29
@mmby @jungkopf @aeva @ylegall yes b-spline approximation is not trivial. this is the tough part.
Show threadmmbyMay 29

@lritter @jungkopf @aeva @ylegall there is one very pointy part of an isoline where that 0 and 7 is

if you carefully watch it, the numbers don't change angle very smoothly, they discretely flip, may not be spline approximated but just linear segments that describe the isolines

Show threadLeonard RitterMay 29

@mmby @jungkopf @aeva @ylegall spanning and rectifying the b-spline itself is closed form, yes, can also just be linear, but grouping and connecting the points is not something you can do in a single pass compute shader. if it were, the isolines could indeed even be animated.

if you built a pseudo-distance field from isolines, snapped random points to the roots and had them flock together (attempt equal spacing along tangent, take same direction of travel), that could work in realtime.