For anyone else struggling like me to translate this to python. Here is a solution:

Oh dear, forget about it. I was on 18.5.499 all the time. Sorry for being stupid, and thanks again for the help

Thanks again for the effort! I'm starting to feel stupid but I still can't get this to work. I apply my material to the "static mesh (instanced)" output of my instancer, correct? I had to rebuild your hda because engine won't allow me to import indie assets with a fx license. Can you try if the attached asset works for you?

Awesome, thank you!

No one? Or is the question too confusing

Can somebody help me with this setup? I can't get this to work. I want to assign a random color to my instances. I'm using the attribute instancer with the unreal_instance attr. set on my points. Then I add an unreal_num_custom_floats integer attribute set to 3 and then three float attributes unreal_per_instance_custom_data0,1,2 set to different values.
Inside Unreal I'm using PerInstanceCustomData in my material assigned to the object I'm instancing. But apparently it doesn't contain any values, it's always black. Any help would be greatly appreciated!

Yes, this drives me crazy as well. In a folder with several 8k images this always means time for a coffee break.

Thank you very much for your help. I'm still struggling to get this to work. I've set the “field to match” to surface in the volumesource. Although I really don't understand why, because by default when sourcing flip surface it's set to density. This whole dops/field thing is sooo confusing.
Another error (I suppose) was that I was sourcing my mask only on the first frame, I've set the activation to constant now. But still it's not behaving like I want it to, which is blending in the effect of surfacetension from one direction. Now the fluid is exploding in a weird way…
Any more help would be fantastic!

Anyone?

Hi!

I want to use mask field of the Gas Surface Tension DOP to control surface tension with an attribute, but whatever I try seems to have no effect. I'm using the volumefromattr sop to create a vdb and try to source that in dops, but no luck. Any help would be greatly appreciated!

Cheers,

Felix

Basically I want exactly the same behavior as the translate SOP but for my own HDA. So when I create my HDA the selected components are put into the group field and their average position used as a pivot. Any ideas? Thanks!

Aaah, ok. Thank you!!!

Ok, this is driving me nuts. Can someone explain to me why the attached scene doesn't work?
Inside the geometry I'm defining a vector attribute and add it to P. If I do it in the same wrangle it works. Defining the attribute in another wrangle doesn't. Using a pointvop however works?

Thanks in advance!

Thanks for the input! My impression is that in Houdini there is a more strict separation between animation and shading steps. So for shading/material assignments giant SOPs with cached Animation are apparently the way to go. Although I still think it's super impractical to transform things or even layout larger scenes this way. So how are Animation scenes organized? I guess 99% are caches from other packages any way so not many people seem to care…

PS: Is it just me, or is subscribing to a topic not working? I never got any email…

Hi!

This is another noob question. I tried reading through the various threads regarding this topic but I can't find a really good answer to this.
In XSI we used to have quite deep hierarchies with several SRT nulls/locators to control the individual parts of an asset. Material assignments were done per object or on groups of objects. Maya works similar.

What's the preferred way in Houdini to deal with large hierarchies/complex Assets, especially in regards to material assignments? There are so many ways to do this, but I suppose there must be a best practice common to most companies?

So far I've found the following ways:

• Classic hierarchy using only subnets, no nulls. Material assignments at object level
  

This seems to be pretty uncommon. I guess because it's hard (impossible) to select the subnets in the viewport to transform things? But it's pretty fast to setup and looks quite clean to me. Since there is no way to group objects (I think?) material assignments would be assigned to each object individually.

• Classic hirarchy with subnets and nulls
  

Here we're using nulls feeding into subnets to control the transformation. This has the advantage that we can see and select the nulls in the viewport to transform things. But it's super cumbersome to setup. We have to create the nulls and wire everything together - the nulls into the subnets and each sub-network input into the nulls - instead of building the hierarchies with a single click. Also things get potentially chaotic if transforms are applied to the subnets instead of the nulls feeding into them.
Materials would be applied similar to above.

These two methods can also be quite confusing with the Show/Ghost/Hide objects mode of the viewport. Although maybe that's just something one hast to get used to.

• Flat hirarchy, only one subnet with nulls for hirachy
  

This seems to be quite common. But to me this seems super confusing. When importing FBX or ABC files i really often end up with gigantic trees like the one below.
To be fair, I only recently discovered the Input/Output mode for the tree view which does help a lot to navigate these networks.

All of these methods have the problem that Houdini doesn't seem to be very comfortable with lots of objects in separate object networks. For example I only recently discovered that you can't instance subnets in the viewport? How would you use such assets for large scene layouts? Or it's not possible to edit the mesh of two geometries at the same time.

• Everything in SOPS using groups and transforms
  

I've seen this in a couple of tutorials. Bring in and merge all the geometry with one or multiple object merges. Or combine all the SOPs directly by using the combine tool from the shelf. Then manually or with partition sop, or with wrangles, or…. put objects into groups so things can easily be selected in the viewport.
Apply materials with material SOPs.
But building a flexible hirarchy here is almost impossible. You would need to build an inverse tree with merge and transform sops. There is no way to select the transform sops in the viewport. You can't move the transforms without the input objects (child compensation). Also transforming things in SOPs is much slower than on the object level i think? So of course for animation this really isn't an option (or is it?). So I think the only time this would be useful is when you're absolutely sure you don't want to transform things anymore in SOPs and don't care about the original hirarchy.

• A combination. Use hirarchy for animation/transforms and a single Geometry for material assignments
  

This would solve some of the problems above, but even more cumbersome to setup. You have one subnet using one of the hierarchical methods above (which one?) and an additional geometry where you merge in the subnet objects and assign materials there. But you constantly would have to switch between the original hierarchy subnet and the flat geometry network if you want to change things. You still have to setup groups to separate the geometry in sops again ( I think the easiest way is to setup attributes in the original hierarchy and use a partition sop to generate the groups based on those?). Material assignment is very annoying (no drag and drop anymore) and also very hard to manage if you have to assign hundreds of materials, it's hard to find which material_sop does what.

• any other way ?
  

I've attached a small example scene. Any pro tips and hints here would be greatly appreciated!

Cheers!

Well that's the same as setting the instance object directly to the box inside the subnet, no? This was a simple test, a real life scenario would be a subnet or digital asset with hundreds of nodes. Is this really not possible?

Hi!

Is there a way to view instances of subnets in the viewport? Mantra renders the instances fine (Redshift unfortunately doesn't) but I can't see them in the viewport.

This is on 16.0.671, attached is a simple scene.

Thanks!

As the title says, is there a way to use expressions in node comments? Would be nice to show node parameter values similar to what the file or group node does automatically.