“`c float radius = chf(“radius”);
float dia = volumevoxeldiameter(0, “height”) / sqrt(3);
int voxel_rad = ceil(radius / dia);
float height_max = f@height;
for(int i = i@ix-voxel_rad; i <= i@ix+voxel_rad; i++) { for(int j = i@iy-voxel_rad; j <= i@iy+voxel_rad; j++) { if(length2(set(i-i@ix, j-i@iy)) < voxel_rad*voxel_rad) { float height = volumeindex(0, "height", set(i, j, 0)); height_max = max(height, height_max); } } }
f@height = height_max; “`
Category: hou-general
Archived post by tokyomegaplex
Archived post by mestela
most of the japanese houdini advent posts are online, like every year, some fried gold entries in there: qiita.com/advent-calendar/2023/houdini
Archived post by mestela
someone shared this in the cgwiki discord, an incredible collection of hips and stuff: https://twitter.com/WhiteDo27114277/status/1726265090117406728
the google doc that explains all the setups is wild, i think half the tricky questions that have been posted here, or interesting papers i’ve seen and thought ‘a houdini implementation would be sexy’, he’s done them all: docs.google.com/document/d/1HBQ6CwiDX0MbK7-nBrLU8hF8H08uPbewZF1z8ybdRw0/edit
Archived post by mestela
I don’t think @lewis.taylor. has shared this here yet, he rerecorded the talk he did for SHUG, lots of great tips here on volumes and general best practices for fx: https://vimeo.com/894363970
Archived post by tokyomegaplex
yeah heh
Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20233912/08/23/fakeinflate.hip
Archived post by nwxo
ok friends i have ported this to ocl, one last step is missing, someone smarter than me needs to solve this…
only updating normals within Ocl is missing, then we could run it on the gpu without the sop loop but with the iterations on the ocl node..
Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20234012/07/23/intersectionPeak_v9_ocl.hipnc
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20234012/07/23/c93002ec7958c654252d3a8743583938.mp4
Archived post by definitely_raining
Just tested in a for loop (nodes) rather then the vex foor loop. It’s much more stable as it samples the updated P,N each step
Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20233012/06/23/image.png
Archived post by definitely_raining
Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20232112/06/23/intersectionPeak_v1.hip
Archived post by lewis.taylor.
yeah, don’t bother
divide with unshared = all the open holes
merge back and fuse and you’re now watertight for vdb
Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20234011/27/23/image.png