Thanks for the detailed reply! I was initially looking for a way to keep everything inside a single USD container just for pipeline cleanliness... I used maya2024-2026, UE5 (with plugins Ornatrix)

Thank you for this information. 

In houdini, it seems there is a way to keep the procedural groom inside the USD file... Mostly to be able to cache the groom guides and apply the animation on the full groom instead of caching the full groom sequence (too heavy).

It is not important to keep everything from the stack in Ornatrix, but if we can extract guide info + full groom assignment on which guide we could build a setup to add simulation on guide in the USD file and the full groom would deform accordingly (probably).

To build USD would need to extract:

-guides full (IDs) (+)

-groomfull (the same IDs, to match guides) (+)

-groom relation with guides (?)

-data density for viewport (?)

Do you think this could be done using Ornatrix?

The main goal is to be able to package everything - the model geometry, lookdev, shading, and the groom - into a single assemblyMain USD asset file.

The idea is to load this master USD as a reference in the scene, and then simply attach cha_animation, simulation cache of the guides to it at the scene for the final render. (maya, vray, maybe EU5 too).  This would completely eliminate the massive file sizes and heavy disk read times associated with caching full groom caches.

Storing a reference to the oxg. yaml file inside the master USD (as you mentioned) sounds like a great way to handle the Ornatrix part of the asset. If the Ornatrix plugin in the target application (like UE5) could read that USD, fetch the . oxg. yaml parameters, and then deform/generate the full groom dynamically based on the external animated guides, it would be the perfect pipeline solution.

Is it currently possible to structure a master USD asset this way with Ornatrix, or would this require implementing new export/import features on your end?

Roman Adamanov

Grooming supervisor - AGORA.studio 

rAdamanov@agora.studio

Got it, I'll create a test scene and attach it to my next message. Thank you!

 Hello, the on-the-fly generation workflow you described for Arnold is exactly the core concept we are aiming for, but integrated within a USD asset structure. An important note about our pipeline: our primary render engine is V-Ray in Maya2024-2026

 I have prepared and attached a basic Maya test scene that replicates our standard studio groom setup.

Feature Coverage:

I have deliberately built this test scene to cover the maximum possible range of Ornatrix features to ensure the new workflow is robust. Inside the setup, I have used: A density map to control and mask hair growth in specific areas. Hair groom deformation driven by displacement map. The branchMode feature. A standard set of core modifiers (Frizz, Clump, Detail, Length, etc.)

Here is a detailed breakdown of our scene logic and how we envision the ideal workflow (the video is attached too)

1. We use the AdoptExternalGuides operator for connect simulation to the groomMain

The Outliner structure is as follows:

"groom_grp" - contains the full Ornatrix procedural groom, emitter_geo, AdoptExternalGuides sim setup, control curves

"geo_01_emitter" - the base geometry (scalp/body)

"geo_01_groom_driver_crv" - the driver curves used to guide the shape and dynamics/simulation of the groom

2. Our Current Publishing Pipeline:

Currently, this setup is published in our pipeline as a single master asset called "groomMain"

However, we separate the components so other departments can work with them in the modeling, shading, lookdev, lighting and animation stages:

The groomMain: we save/publish the entire "groom_grp" (with the full Ornatrix stack + emitter_geo + AdoptExternalGuides sim setup).

The Driver Curves (output_curves): Published separately so the CFX department can simulate/attach hair simulation caches to these specific curves directly in the render scene.

3. The Desired USD Workflow

We would love to translate this logic into a strict USD pipeline. Here is a step-by-step of how this should ideally work:

Step A: Export

We export the "groomMain" asset into a MasterCharacter USD file. Alongside it, the procedural groom stack is automatically saved as an external .oxg.yaml file.

Step B: Assembly (The Master USD) During the scene assembly stage (in Maya), we load this master USD. We then connect the character's body animation pointCache to geo_01_emitter and the CFX hair simulation cache to the geo_01_groom_driver_crv curves (via USD overrides/payloads).

Step C: On-the-fly Generation in V-Ray / Maya... This is where we need the new feature. When the assembled scene is opened/imported in Maya for the final render in V-Ray, the Ornatrix plugin should:

1 Read the metadata or a reference inside the USD pointing to the external .oxg.yaml file.

2 Reconstruct the procedural modifier stack.

3 Automatically pick up the incoming USD animation from the emitter (geo_01_emitter) and the simulation deformation from the curves (geo_01_groom_driver_crv), feeding this data into the AdoptExternalGuides node.

4 Dynamically generate the final, high-res animated hair on the fly during rendering with V-Ray.

Please let me know if this breakdown makes sense, or if I need to adjust the test scene in any way to make it easier for you to test.

Thank you again for your time and support!

Hello, I think it should works on our side... Exporting our hair/groom result in a USD format that we can still drive with guides. (USD referencing to an external .oxg should be ok). I think we would be using native mayaUSD