instance simulated geo

[hatrick]

hi,

I revitalized an old houdini experiment... i try to simulate a cloud of slowly falling feathers. i ve roughly setup the feather with fur and added dynamics to the structure...everything works quite good by adding the skingeo into a dop network first...

Now my question is how i can instance the allready simulated feathergeo(wiresolver) for example on a particle system ore an initial template geometry without loosing the wiresim of the single feathers ... goal is a layered simulation with different feathers of different complexity.

thx a lot

Edited September 15, 2010 by hatrick

[hatrick]

something like this

[pclaes]

So to simplify, you have a bunch of animated caches sitting on disk and you want them instanced on points?

Have a look here:

http://www.peterclaes.be/blog/?p=34#more-34

[hatrick]

So to simplify, you have a bunch of animated caches sitting on disk and you want them instanced on points?

Have a look here:

http://www.peterclaes.be/blog/?p=34#more-34

as you ve stated caches ... how i can cach such a feather obj and preserve the render behavour of the fur geo with all the different control shaders plus the skin geo?

thx

[pclaes]

ahhh.. I see. You can't if you are using fur.

The fur is already a VRAYprocedural so it is generated at rendertime. You can't have an instancer call upon the fur again unless if you were to write your own custom VRAYprocedural, but that would not be an easy task and you would have to program it in the HDK.

Therefore:

I don't know if it is possible to make a level of detail versions of your feather(s)? After all fur is only curves and the shaders applied to it are CVEX/Vop networks. So can you bake a feather out to disk that looks the same as it would at rendertime (Like a 100% representation of a single feather in your viewport?). I haven't used fur enough for that, but as I said, it is just curves at the end of the day.

The closest thing I can think of to "catch" the geometry is with a pcwrite() added to your shader at rendertime. It would write out a pointcloud that would represent the feather based on the micropolygons. But that seems convoluted and rough.

Personally I would model a single feather procedurally in sops so you can do whatever you want. The fur that is meant to be used with feathers seems more like a "textured plane" approach.

[hatrick]

ahhh.. I see. You can't if you are using fur.

The fur is already a VRAYprocedural so it is generated at rendertime. You can't have an instancer call upon the fur again unless if you were to write your own custom VRAYprocedural, but that would not be an easy task and you would have to program it in the HDK.

Therefore:

I don't know if it is possible to make a level of detail versions of your feather(s)? After all fur is only curves and the shaders applied to it are CVEX/Vop networks. So can you bake a feather out to disk that looks the same as it would at rendertime (Like a 100% representation of a single feather in your viewport?). I haven't used fur enough for that, but as I said, it is just curves at the end of the day.

The closest thing I can think of to "catch" the geometry is with a pcwrite() added to your shader at rendertime. It would write out a pointcloud that would represent the feather based on the micropolygons. But that seems convoluted and rough.

Personally I would model a single feather procedurally in sops so you can do whatever you want. The fur that is meant to be used with feathers seems more like a "textured plane" approach.

thx peter

[pclaes]

Just having fun ...

instance_feather_01.hip

[hatrick]

Just having fun ...

wow many thx peter... but one further very common question about simulation... is it possible to usw both pops and dops for oe sim. let's say i'd like to emit complex objects via pop's instancig which should behave in terms of coolission like simulated in a dop network...

greets

[pclaes]

instancing and "behave in terms of collisions" in dop network don't really mix.

Sure you can have a bunch of particles colliding with dop objects (you can probable get a spherical influence radius, but no complex collisions as you are dealing with a single point), but the instancing process happens at rendertime. The geometry is instanced onto a point and has no knowledge of any collisions.

So if you need some hero feathers colliding you will probably have to sim them. I would recommend simplifying the geometry to a simple grid with a few divisions that you can turn into a cloth object (cloth solver will try to solve inter cloth collisions). Or perhaps into a squishy object (particle/spring based solver, I doubt this will allow inter-squishy object collisions). The actual feather curves can be wrap deformed (with either a lattice (gives rotational problems) or a meshcapture-> have a look at the cloth masterclass where that low res car deforms a high res car mesh, there is an asset in there that will help with that wrap deformation.)

In terms of automating some of the dop setup I would recommend having a look at how the fractured geometry in RBD's is set up.

This is were fur as fur shines as it generates the curves at rendertime. Depends how detailed a collision you want, if the individual strands of the feather will need to collide you will need to use guide curves, but that would probably be for one or two close up here feathers.