Hi
I need to create a large crowd of 50,000 people from an ancient civilization, and the characters are all wearing long robes. I created the outfits with Vellum drape and cloth, and plugged the cloth solver into the crowd solver and then realized that it will run a cloth sim for every crowd agent. 50,000 cloth sims - that is impossible on my hardware. Is there a way to sim and render out a few cloth animation clips and use that when populating the agents, as opposed to a full cloth sim for every single agent?

Any help would be tremendous!!

You could try to convert a pre-baked cloth sim to skinned geometry (e.g. the Labs skinning converter tool). Or, depending on the shot you could get away with only running cloth sims for foreground agents?

technically you could also utilize the new agent blendshape support too, however I haven't used it yet so I'm not sure how optimized it'd be for such amount of blendshapes

The agent blendshape deformer is very similar to the Blendshapes SOP, so it would be a bit awkward for this purpose (you'd need a lot of channels, whereas for this you'd want a deformer that behaves more like the Sequence Blend SOP)

cwhite
The agent blendshape deformer is very similar to the Blendshapes SOP, so it would be a bit awkward for this purpose (you'd need a lot of channels, whereas for this you'd want a deformer that behaves more like the Sequence Blend SOP)

yup, a lot of channels, definitely not ideal, but a technical option to pack exact deformation into crowd clips

Thanks you so much for all the help guys. I really appreciate it.
After many days of experimenting with various options, the closest I got to getting something that is usable is to bake out the cloth sim to an MDD file, and then use that as a agent layer. However, when I try to bake the agent and the MDD layer, with a ROP node, it only render's the “create-agent” or base geo, and only the first frame of the MDD sequence. So I have this suspended clothing that the character walks through. What am I doing wrong? Is there a specific way to bake the agents with the MDD animation layer?
I suppose I can use the MDD layer in the crowd sim without baking the agents, but then I cannot randomise the clip time in the crowd solver. For the closer shots this might be more of a problem, but most of the shots are drone shots. Any insight would really go a long way.
Thanks again everyone.

well, no, packed agents don't support any sort of point cache
you only have 2 options, as mentioned above, either bake your deforming geo into skinned geo with set of bones and corresponding animation channels, or you can try using new blendshapes support, which would entail baking the whole sim into blendshape per frame and channel per blendshape where it turns on for that frame

Blendhsapes are awesome. Thanks so much Thomas and cwhite! Legend! Happy 2020 everyone.

tamte
well, no, packed agents don't support any sort of point cache
you only have 2 options, as mentioned above, either bake your deforming geo into skinned geo with set of bones and corresponding animation channels, or you can try using new blendshapes support, which would entail baking the whole sim into blendshape per frame and channel per blendshape where it turns on for that frame

I'm bumping this as I'm running into this same issue and I couldn't find anything more recent on this topic.
Is the situation still the same (H20.5.487)? No support for any sort of cache on agents (or agent layers, for that matter)?