Forums

I really like the cellular noise node that can be used inside a point VOP, as it yields good borders compared to trying to remap worley/voronoi patterns. The downside is that it can only be used on a 2D uv map. How would one go about getting the same results with a vector 3/3D input? I've tried searching, but haven't really found anything directly addressing the issue. Thanks!

Hi; I have an asset that I imported into Houdini that has pieces with reverse Vertex Normals inside a for each loop per connected piece node, I have written a vex code that finds their bbox center of a piece and tries to determine which pieces get flipped according to the dot product of the point positions and the the bbox center point and have gotten no result out of it And this is what I could come up with for islands with volume. I have absolutely no idea as to how to handle flat plane like structures. Maybe compare their vertex normal orientations to the centroid of the entire enclosure? But then again my code doesn't do the job Any help would be appreciated. AJ flip_reverse_prims.hiplc

Hi Daniel, It's 6 years later and I've been looking into doing the same thing. Any possibility in sharing please? Thanks!

thanks this fixed my issue!

https://www.patreon.com/posts/polar-inversion-159637654 I have been experimenting with Houdini’s new Copernicus network and started porting some of my existing COP tools into custom OpenCL COP nodes. The first one is a Polar Inversion filter implemented directly as an OpenCL COP node. The core mapping is based on radial inversion around a center point: scale = radius² / distance² In practice, the important part is not the formula itself, but making it stable and usable as an image filter. The implementation uses inverse mapping, where each output pixel traces back to the corresponding source position. Near the inversion center, the denominator is clamped using a pixel-footprint floor, with an optional softness floor. This avoids the unstable singularity you would otherwise get from dividing by a near-zero distance. For sampling, the node uses deterministic subpixel supersampling and Houdini’s rectangular filtered sampling through "imageSampleRect". The filter footprint is currently based on the output pixel footprint, which gives better reconstruction than a naive point lookup while still remaining predictable and fast. This is the direction I want to take with Copernicus: converting my existing COP nodes and implementing more advanced image filters, compositing operators, and procedural transformations that are not currently available in Houdini. The goal is to build a deeper image-processing toolset inside Houdini, using the same philosophy I use in my VEX work: low-level control, practical performance, and tools designed around real production needs.