How to render smoothed particle hydrodynamics (SPH)?

[Ken]

Has anyone used Indigo for rendering fluids and materials simulated with smoothed particle hydrodynamics.
What is the state-of-the-art method for doing it, and is there a recommended pipeline for it?

Assume I have a program that simulates a fluid with SPH and can dump the usual SPH data, e.g. particle positions, velocities and densities. Now I want to render this in the prettiest possible way (hopefully also efficiently). Should I compute a surface mesh using e.g. a marching method, or should I use particle kernels to compute iso-surfaces directly for the light model (e.g. using blobs or similar)? Are there methods for this in or around Indigo, and/or should I go via a modeling tool, e.g. Blender or Maya?

Is anyone using RealFlow for fluid simulation and Indigo for the rendering?

[Borgleader]

I think zsouthboy wanted to try using realflow for fluidsim but dropped the project.

I didn't see many fluid sim projects around the place. So it's hard to say.

I do know you can use blender's internal fluid sim with indigo (since it essentially becomes a mesh like any other).

[alex22]

There is actually a video on Youtube showing fluidrendering out of Blender: http://www.youtube.com/watch?v=83I4grJ0 ... re=related

[Zom-B]

Ono did some experiment over 1 year ago with 3d density volume datasets:
http://www.indigorenderer.com/joomla/fo ... 3883#23883

[Ken]

There is actually a video on Youtube showing fluidrendering out of Blender: http://www.youtube.com/watch?v=83I4grJ0 ... re=related

Yes, but as far as I know this is using the Lattice-Boltzmann method in Blender, and is thus grid based, so this is a pretty different problem compared to rendering a particle based fluid.

[Borgleader]

There is actually a video on Youtube showing fluidrendering out of Blender: http://www.youtube.com/watch?v=83I4grJ0 ... re=related

Holy shit that is epic. If only I knew how to use the fluid simulator to do that

[Ken]

Ono did some experiment over 1 year ago with 3d density volume datasets:
http://www.indigorenderer.com/joomla/fo ... 3883#23883

So, what you suggest is that I sample the SPH density field into a uniform 3D grid and use the same method? This could be done (and I have also tested this in povray a few years ago), but I was hoping someone had worked on a more direct method for rendering particle fluids in Indigo.
It should be possible to optimize a metaball type of approach quite alot, and in this way avoid storing all that 3D-data.

I'm a total newbie on Indigo, so sorry for asking newbie questions, but does Indigo support any type of metaball or blob rendering primitives, or does it have support for creating isosurfaces (e.g. marching cubes)?

[StompinTom]

AFAIK nope... its all mesh-based right now i believe. though looking at some of the experiments Ono has done in the past, i wouldnt be surprised if hes got an experimental version of Indigo that hes keeping all to himself, the greedy bastard

[zsouthboy]

I think zsouthboy wanted to try using realflow for fluidsim but dropped the project.

I would've ended up with triangles out of there - not what he's looking for.

So no, don't look for help from me - I've got no idea.

Might want to talk to Ono since, as others have said, that he did at one point test using 3d data directly in indigo (smoke)...

[CTZn]

Hey Ken don't ask me to go in details but you can try to instance a sphere primitive (parametric indigo primitive) shaded with sss or just absorbance for each particle. This is theory of course

[Ken]

I think zsouthboy wanted to try using realflow for fluidsim but dropped the project.

I would've ended up with triangles out of there - not what he's looking for.

So no, don't look for help from me - I've got no idea.

Might want to talk to Ono since, as others have said, that he did at one point test using 3d data directly in indigo (smoke)...

Well, triangles is fine as long as it just works and gives cutting edge results.

I think it remains an open question how particle materials should be visualized, e.g. using parametric surfaces from kernel functions, using meshing techniques, using splatting techniques with oriented disks, or using raw 3D sampled voxel data.

Though this is an incredibly interesting and important subject for research, both for high quality rendering, and for interactive rendering, I am right now just looking for a workflow that really works and gives really good results.

I have my own simulations and can thus output them in any format suitable for a rendering engine. I'm not after full producution workflow, but something that makes it possible to visualize research results on new and improved particle methods.

I have tried using Povray's metaballs and while this gives really nice results, it is horribly slow if I have more than a few thousand particles, so I would have to do some sort of surface detection on this data and remove the excessive bulk information. On the other hand, that makes it impossible or at least very difficult to have bubbles and vortices.

[CTZn]

I think it remains an open question how particle materials should be visualized, e.g. using parametric surfaces from kernel functions, using meshing techniques, using splatting techniques with oriented disks, or using raw 3D sampled voxel data.

or procedural speres with same material I think they are implemented in blendigo, could be done for Maya too. But I didn't test that... but that should be faster than many techniques you mentionned since sphere primitives have 2 parameters; position and radius.

[neo0.]

Its a pity that the only fluid simulation is in blender, since many peopel find blender's UI to be confusing.

Maybe indigo could have it's own fluid sim? A lot of the stuff is already free
http://elbeem.sourceforge.net/

[CTZn]

Fluid sim is time dependant, Indigo is not, yet. Wrong way neo0. It's too bad you dind't try blender harder, you said.

[dougal2]

Its a pity that the only fluid simulation is in blender

erm... not quite. Maya has had full dynamics+fluids since before blender