Hi I am trying to carve a curve at corner points...
the goal is make a bend with a set inside radius

ok so I cut out a straight line from a corner so a bit like a chamfer but it is sized using some vex and carved to size by the geo's width and what ever size internal radius is wanted. And I basically create a three point ark.

for this to work the length from the corner point needs to be cut equally ether side or the ark will be bumped off line or not blend in well leaving a sharp edge allowing geo to overlap again.

The PROBLEM comes when bends are too close to each other not allowing enough geo on one side for the correct sizing.

Originally I was thinking to split the failed bends and do a second pass but this was not the fix. so now I believe that I need too carve the tightest bend first re-check angles to see if any nearby registered bends are still there or has the first carve cut it away and do the next bend until all required bends are made. I was using for each named prim before but the new idea has made my brain leave the building as it is trying to work out how to do all this in one wrangle node.

Hello

I'm wondering if the "dual LPE" syntax described in the OSL docs is supported in Karma?

OSL Light Path Expressions [github.com]

Good, you got your alpha channel as a separate AOV, but what if you want to put it in the actual alpha channel of an RGBA AOV? You can do that by appending your expression to the standard rgb expression for an AOV. Look at this example:

Output ("beauty.tif", "rgba", "C[SGs]*D*<Ts>*L !C<Ts>*B")

初心者です、紙吹雪をテストで製作していました
地面に接した時に地面のノーマル方向に角度を変える（突き刺さらないようにする）挙動をやってみたかったんですが


上記popwrangleをpostsolverに繋ぐと最初からY軸方向を向き続けます
geometry Spread sheetを見てもhitnumの挙動は正しいように見えるんですが、全フレームでif文が発火してしまっているように見えます
根本的な所で勘違いをしていると思いフォーラムに投げさせて頂きました

お手数おかけしますが、ご助言いただけると幸いです

Ongoing issue with support but basically in most of my test having default caching enabled on pyro block end makes it completely unusable. Eg live sim or flipbook with caching disabled 150 frames approx 8-10 min. Turn on caching (the default) and it will take 30 min just to get to frame 50.

This is on windows 11 (all updated) with latest studio nvidia drivers and 3090 RTX 24GB

So to get the speed as weird as it sounds - disable caching on the pyro block end when working with pyro.

Hey guys,

I've recently ran into a set of GPU-related crashes when trying Copernicus (File, Constant COPs) and OpenCL SOP nodes. It looks like, though not a solution, there's a workaround, which hopefully may be solved in the future if SideFX decides to support Intel GPUs:

Add the following 2 lines into your Houdini start script:


I hope that may help somebody in the future.

Cheers!

NOT SOLVED YET :
the solution i found , not works if the previus asset has scale transformations too.
I will make another post flagging this error , with better info and explanation.
If you check on the toogle "local transform" , it fix the problem

Short Summary

Stage Manager becomes unusable if assets already have transforms. This makes it impossible to use Stage Manager reliably in layout and set dressing, breaking the most common USD workflows. The problem has existed since 20.5 and is still present today.

Detailed Explanation
Why this matters

Solaris is designed to be a complete USD workflow, from asset creation to final render. But right now, the Stage Manager — which should be the central node for layout and set dressing — only works once. As soon as you try to reuse it downstream on assets with pre-existing transforms, pivots and transforms break.

This is a show-stopper for production pipelines. Both small and large studios hesitate to adopt Solaris as their base pipeline because of this limitation. Without a reliable Stage Manager, teams are forced into awkward workarounds (dressing in SOPs with point imports, or building in-house tools) instead of using Solaris as intended.

Steps to reproduce (3 simple cases)

Case 1

Create a test rubbertoy in Solaris.

Apply scale/rotation/translation.

Add Stage Manager → move or duplicate → transforms break.

Edit LOP works correctly in the same situation.

Case 2

Create rubbertoy.

Add Stage Manager → move/duplicate/rotate.

Add another Stage Manager → pivots/positions become corrupted again.

Edit LOP still behaves correctly.

Case 3 (real production scenario)

Layout department publishes blocking with Stage Manager (placing assets, variants, transforms).

Set dressing department loads that opinion and tries to refine or duplicate assets.

Stage Manager fails — transforms break because layout already added transforms.

Result: impossible to refine layout with Stage Manager in set dressing.

Why fixing this is critical

Stage Manager should be the main tool for layout and set dressing in Solaris.

Without it, Solaris cannot be used as a full USD pipeline.

Forcing workarounds discourages adoption: some studios move dressing to SOPs, others avoid Solaris altogether and stick with Maya.

Fixing Stage Manager would unlock Solaris as a robust USD pipeline tool, not just a lookdev and rendering playground.


Attachments

.hip file demonstrating the bug

Video showing the issue step by step



Do you also encounter this issue in production? I’d love to hear your experiences — the more visibility this gets, the higher the chance of a fix.

Thanks!!!!
best regards!!

Sometimes you need to deform an object inside another with respect to the internal volume, for example muscle inside a body.


One way is converting the object to tetrahedrons, since tets have 3D primuvw coordinates for interpolation. Sadly this isn't always an option, because the triangulation changes when animated.

A better approach is using Positive General Mean Value Coordinates (PGMVC), as implemented many years ago by Sibbarick [forums.odforce.net].


It works by casting rays in a sphere from each point, creating lines. Each line acts as a distance constraint, pulling the point as the walls move. It's almost like Point Deform with occlusion, where only the faces visible to a point light are used for deformation.


I rewrote Sibbarick's method using primuv to make it simpler, with slightly better results on the exterior.

Download here:
https://github.com/MysteryPancake/Houdini-Fun/tree/main?tab=readme-ov-file#hda-volumetric-deform [github.com]

Hi All! I'm working in H20.5 (can't wait to get to H21 but I'm on MacOS M-Series systems and it'll be a while before Otoy gets around to getting a H21 build for me . I have just a simple question that seems like it should have a very simple and obvious answer but I'm struggling. So, I have a very simple APEX humanoid rig with FK and IK and controls. I would like to add noise to a keyframe animation (or the complete animation) via CHOPS nodes. But I just can't figure out how. I've tried many approaches. Any help would be greatly appreciated!!!


Thanks,
Greg

I have first node with 1000 points (no more attributes)
I have second node with 2000 points (no more attributes)

And I change some of these points in std::thread

And I have third node - Merge - first + second

How I can receive these points from Merge node? HAPI say, what there are 0 parts and 0 geos in Merge node.
How Merge Node saving all points and how I can get it in HAPI?

Houdini Session showing correct values in Geometry SpreadSheet. How I can show such points table in my own app?

Is anyone having trouble getting material instance parameters working? Wondering if this is a user error or a bug.

I'm following the documentation, but can't get the "unreal_material_parameter_diffuseboost" param values to transfer over like in the example. My material instance is being created fine, however.

https://www.sidefx.com/docs/houdini/unreal/materials.html#instances [www.sidefx.com]

Currently I have a tiny hqueue farm setup on 3 machines on Windows 11. Both the server and all clients are on h20.5.332. Whenever I decide to upgrade to h21, do I need to upgrade the clients and server to be compatible with h21 jobs?

We're looking for a senior Houdini (or Unity!) dev to join our Home Tee Hero golf simulator team.

While this job is classified as a Senior Software Engineer and we are looking for a solid technical background, the role is primarily Houdini/Unity focused, and a great portfolio goes a long way to bending the technical requirements.

This is an on-site, full-time position, in Kansas City or Raleigh, NC.

https://careers.garmin.com/jobs/15105 [careers.garmin.com]

Hello Everyone,

I'm Namrata Hinduja Geneva, Switzerland (Swiss) and I'm exploring procedural environment generation in Houdini and aiming to create a system that dynamically adjusts terrain features (like erosion, vegetation density, and object scattering) based on user-defined gameplay parameters (e.g., difficulty level or biome type). While I've had success using VEX and SOPs for isolated features, I'm struggling to build a cohesive system where multiple layers interact responsively. Is there a recommended approach for managing such interdependencies—perhaps using TOPs, Attribute Wrangle workflows, or custom Python tools? I’d love to hear how others structure complex procedural systems that need to be both artist-friendly and scalable across large environments. Any insights or references are appreciated!

Regards
Namrata Hinduja Geneva, Switzerland

We've been experiencing a weird issue when loading a bunch of USD assets onto a scene that contain a usd preview surface material set to preview purpose, where the viewport is constantly drawing and seems to be hung up on the USD Stage Material setup. It's been difficult to pinpoint what triggers this, at the moment it's random and sometimes switching desktops or labs resetting the viewport fixes the problem, on other times we just have to close the scene.

This gets really problematic when we have a scene with 100+ assets, turning textures off on viewport doesn't help. The only other solution is to drop a unassign material LOP node. Once this issue happens even if I try to reference in a single asset the same thing happens, So I don't think its anything specific to having an assembly of usd assets.

Could this have anything to do with the assets having textures on a network drive? Unfortunately can't share a scene because of local studio assets but can attach a screenshot of the way all of the assets are structured.

If anyone else has experienced this I would love to hear your input!

not sure if this is a limiation of some sort (from exr), but I am unable to write multi aov exrs with generate mipmaps. if I try so, the rop only writes the first aov (which will be tiled) but none of the others.

The H21 daily builds are now online! Enjoy!

Hello,
I am curious if it is possible to set the initial state using vex when setting up crowds? In the image I just have things set to random distribution between 2 walk states on my crowd source. I want to be able to set initial states based on other starting variables and attributes, such as free space around the agent point (if lots of space = big walk, if little space = small walk).

I tried just a point wrangle after the crowd source node to 'override' the state, which in the goemetry spreadsheet does look like it updates. But both visually, that agent point stays stuck in the state it was assigned inside the crowd source sop, and the sim ignores my wrangle as well.

I am working around this by making my agents start in an idle state and then having some crowd triggers to either get them to go into a big walk or a small walk. This is ok, but I am starting to get a pretty messy network having to have 2 separate walk states and all the pop nodes connected to them, so I am not sure this is a good workflow. Any help is much appreciated!

Hello, I'm following this tutorial series, and I'm having trouble at 0:57. I tried following one of the commentor's suggestion to "Enable Deprecated Instance Variations" but I don't see it in my settings. How do I enable this feature? I'm using UE 5.5.4, Houdini 21.0.440.




https://www.sidefx.com/tutorials/foundations-procedural-assets-for-unreal/ [www.sidefx.com]

In general, when we want to generate and save an asset using the Component Builder, after clicking *Save to Disk*, a thumbnail is created inside the asset’s folder.
The issue is that the two files `.usda` and `.png` are saved with the generic name **thumbnail**. Then, when we try to load these assets via *Add Asset Directory*, since the thumbnails weren’t saved with the asset’s actual name, the system generates new thumbnails again for those assets.

This becomes a problem if we had already created our own custom thumbnail (for example, with a specific camera angle or lighting setup), because we can’t use that custom thumbnail in the Asset Gallery.

Interestingly, this issue does not occur when the asset has variations — in that case, all thumbnails are saved with their respective variation names. However, this problem remains for single (non-variation) assets.

To solve this, I went into the **Component Output** node and changed the **Override Output Image** setting in *Thumbnail Render*, as well as the **Output File** option in *Save Thumbnail Scene*. Instead of saving them as `"thumbnail.png"` and `"thumbnail.usda"`, I replaced them with:

```
chs("../rootprim") + "_Thumbnail.png"
chs("../rootprim") + "_Thumbnail.usda"
```

This way, both files are saved with the asset’s actual name, preventing issues when loading them back into the Asset Gallery.

I just wanted to report this so that the dev team could make a small fix for it.

I tried baking high-poly details onto the low-poly UVs. Since the low-poly model was generated from the high-poly model using polyrebuild, I used the UVmatch mode.
While baking the Height map, I noticed that even the concave areas resulted in positive values.
Did I miss some settings?