Liesbeth Levick

It's not working for you because you bypassed the Guide Reduce which is required to find one curve per clump. As soon as I enabled it again, the glue constraints were generated. This is on your main hairs. The result of the Guide Reduce in this case, for the clumpid generation, would only be used as an input to the Guide Deform which finds the weights for each clump curve.

The Guide Reduce should be able to detect a more random clump distribution if it exists in the incoming groom. It tries to find one curve to best describe all the surrounding curves, and it isn't density based - it is just based on the Target Guide Count parameter. However, you can make use of the Density Mask override on the Guide Reduce to force it to pick fewer clumps in one area compared to another.

gregjenings
Finally, I have one more important question. If this method works well, how would you recommend simulating and rendering millions of hairs? Simulating all the hair directly seems to require a lot of memory.

Is there a more efficient or memory-friendly workflow for handling millions of hairs while still getting a good simulation result? For example, should only a smaller number of hairs be simulated and then used to drive the rest?

Yes, you would simulate a smaller number of hairs and drive the full groom with the simulated ones. Take a look at the "Capybara Sim and Deform" recipe on the "Hair Utils" shelf to see how to do it. In it, a Guide Reduce is used to create a subset of guides to simulate. You might find that the Guide Deform using a very sparse set of simulation guides gives good enough results without needing to set up Glue Constraints on a denser set of simulation guides. Turning on "Preserve Clumps" on your final Guide Deform can also help. If you do choose to go the Glue Constraints Route, you will end up with one Guide Reduce to create your simulation guides, plus another Guide Reduce to calculate the clump centres of those simulation guides. You will also use a total of three Guide Deform nodes. One to capture the clumpid, one to deform the simulation guides by the skin, and one to deform the renderable hairs by the simulation guides.

I've added this setup to your file. The yellow nodes are the nodes from the recipe, while the purple ones are from the clumpid detection and glue constraint creation setup.

Please submit a bug to support, along with details of if you have experienced the same issue in earlier versions of Houdini or not. It would also be helpful if you included a file with your submission.

Glad I could help.

Here is a solution Kai and I came up with to your problem. It does a pretty good job of finding clumps.

1. Use a Guide Reduce SOP on your groom. The "Target Guide Count" parm will control the number of clumps it finds.
   
2. Follow this with a (H22) Guide Deform SOP with "Mode" set to Capture, and "Max Guides" set to 1. The first input will be your full groom and the fourth input will be the clump curves from the Guide Reduce.
   
3. To convert the captured guides to clumpid, follow this with a Primitive Wrangle with the expression set to

   i@clumpid = i[]@guides[0];
   

I've attached a file with a simple example to demonstrate.