nuki

length = sqrt(x²+y²+z²) the way i read it, length2 = x²+y²+z² in some cases (eg. when just calculating the ratio between two distances) you can simply ignore the square root calculation to increase your codes performance.

have you tried the sidefx labs quad remesher (requires license) or the instant meshes (open source) bridge?

for deforming geometry use the pointdeform sop. if you have multiple objects this is a great thread on setting it up and maximizing performance. transformpieces is for matching packed/rigid stuff (eg. the output of a bullet sim) . pointdeform.hiplc

if you used the package method - I had trouble to get houdini to load the package until I set the QLIB variable in the .json file to an absolute path. Maybe its a Windows+OneDrive thing.

if you need proper collisions use pop grains like the others suggested if you want to try to use your flip sim I've attempted to use a pretty soft pointrelax (to keep the jitter low, although it should be fairly stable when you do it in your flip dopnet with a few substeps) and a for loop to maximize pscale without intersections. That way pscale is determined by the points distance to its closest neighbour and can change over time (because 'collisions' are soft). Idk, maybe does the trick for you or not soft_deIntersect.hiplc

based on the mapbox documentation, those DEM files are pngs with 0.1 mete height increments. I'd suspect that the texture resolution is lower than the tesselation on the terrain making those height steps visible. So you'll probably need some type of filtering like j00ey suggested.

Is the dem data 16/32 bit float? what range of values are shown in the picture? do you convert/process the dem data somehow?

sure - that just changes the name filter for the output objects from * meaning "everything" to "anything starting with vellum". In this case it isolates the pig heads, because they are emitted into vellumobject1, just to get rid of the groundplane. You could achieve the same with a group and a blast or split sop for example.

like this? V_test_02.hiplc

Its not the most beautiful solution, but seems to work: 2Dbullet.hiplc would be interesting to see how well this works with constraints. To fix the rotations for a "true" 2D look I'd have to decompose the packedfulltransform. I'll have a look at that tomorrow.

hey, i just recreated the thing with a foreach loop and bend deform er in h18. You just need to mark the columns with a "class" prim attrib (eg. with the connectivity sop) deform.hipnc

No issue, glad if I could help! If you want to art direct the effect you can enable the outer glue constraints inside the dopnet and delete them manually inside the "break_cons" sop solver (to control the forming of cracks). You'd do this by adding the constraints to the "broken" prim group (eg. inside a wrangle by setting i@group_broken = 1;). Theres already a few in there breaking the soft constraints by their strech and breaking glue constraints based on force. You could also make them stronger eg. in the Y direction using the dot product of the constraint direction and the y axis. That way the cracks could form more vertically.

Hey there, sorry for being so late...busy week. I managed to get a decent result with a shaking force and a combination of soft constraints (for oscillation) and hard constraints (for clustering). This was a useful reference. earthquake2.hiplc

You could glue your object to a shaking plane or use a custom force with a distance falloff (or a mixture of both). Theres plenty graphs documenting the elevation/shear characteristics of real/synthetic earthquakes acceleration forces over time so that might be a good starting point. If you can recreate the curve it should be easy to drive a force or position with that - I'll see if I have time in the evening since its an interesting topic.

nice job @toadstorm! I was looking into this myself but my solution was about transporting N and up with the sim. Was wondering how to do it with quaternions!