A glass of wine, not cranberry juice

[Macha]

I re-admired Ratatouille the other week and I am always impressed by how precise, stable and beautiful Pixar's effect work is.

The wineglasses were especially lovely so I tried my own version in Houdini.

As you can see there is a problem with the surface not sticking properly to the glass. There are wavy pockets of air that create a weird effect. Any ideas on how to eliminate this?

[sam.h]

Are you using the feature on the particle surfacer that lets you give it a collision SDF to cut away the surface?

Maybe just make the make the particle radius bigger and use that SDF cutty thing (with a high res sdf) to smooth it out ... ?

[mgt]

i am trying something similar, although my glass is more like a 10 sided thumbler. anyway, how i did it was to use the cookie node to subtract the glass geo from the particle fluid surface. as long as i have the particle fluid surface intersecting thru the glass. also turned of pre-convex on the cookie node. then used a transform node one the fluid surface to scale it a tiny bit, like 0.995.

tried the subtract collision volumes on the particle fluid surface node, and it takes way too long for a beast of a computer to bake a high rez isooffset of the glass, and doesn't look nearly as good.

again. this may have worked for me since i am using a sided glass. cookie node is pretty unpredictable.

i can't believe i still can't model a goddamn glass that will refract properly...

[Macha]

Hm, this is tricker than I thought. The problem is not the shape of the sdf but the voxel resolution of the fluid. Because the fluid is round I get this sliced look.

Appending a smooth sop makes the whole thing look too smooth.

Subtracting collision volumes is an idea but I have imported this from the surface/Visualization DOP field and I'm not sure how to reconstruct a good surface out of this if I use particlefluid sop on it.

I tried a lattice deformer set on points. That pulls the fluid close to the glass walls but retains the waves

Attribute transfer the P of the wall edges has some effect but, it messes up my normals and I haven't found a good way to handle that.

Hmmmm!

[tjeeds]

Haven't actually tried it but how about using a ray sop to snap the non-upward facing fluid points to the glass or setting up a VOPSOP using intersect?

If you're careful with the Spread Angle while grouping the points you might even be able to hack a meniscus

Edited February 26, 2010 by tjeeds

[sam.h]

Haven't actually tried it but how about using a ray sop to snap the non-upward facing fluid points to the glass or setting up a VOPSOP using intersect?

If you're careful with the Spread Angle while grouping the points you might even be able to hack a meniscus

That is a great idea

<3 ray SOP

[Macha]

Thanks, I tried it but I cannot make a smooth blend between those points that were selected and those that weren't.

Ahh I feel stupid. Maybe somebody is cleverer than me. I attached a simplified file of the problem.

Examplefluid.zip

[Macha]

Actually, this works not too badly. If somebody can improve on it, let me know!

examplesmooth.hipnc

Edited February 26, 2010 by Macha

[Macha]

image of smoothed problem surface

[Macha]

This is one frame. It looks much cleaner now. I'll render that over the weekend and hope for the best.

Meanwhile I will think of ways to include bubbles. Perhaps a few pop balls advected by the vel field?

[Macha]

Here is the newly animated version:

The fluid is much better but still not good enough. There is also a weird flicker in the glass reflection in the first few frames. Very inexplicable.

[Macha]

Today I added some bubbles. They look and animate OK but I need more work on the fluid. It is too syrupy and slow. I think the flicker is a prob with the PBR render so I switch over to micropoly or raytrace for a while.

Edited March 1, 2010 by Macha

[mightcouldb1]

I agree with you that it's kind of syrupy...

How many particles are you simulating in the glass? Do you have high viscosity/surface tension?

[Macha]

How many particles are you simulating in the glass?

I use volume fluids for this. I don't like particle fluids much because I find them very hard to upscale and control.

I lowered the viscosity today and added some bubbles. Switching off PBR also helped with the flicker in the reflections.

It still looks a bit too thick, but I think that has a lot to do with the bubble animations and I think I can solve that. They are much more stable than I dared to hope for!

A bit of careful design could also help taking care of the glueyness.

Edited March 2, 2010 by Macha

[edward]

Attribute transfer the P of the wall edges has some effect but, it messes up my normals and I haven't found a good way to handle that.

You can always just append a Facet SOP and turn on one of the compute normal toggles. I think the Ray SOP will probably work faster but you'd still want to recompute normals in that case anyhow.

[Macha]

I am now working on speeding up the render, without loosing too much of the look. The frame below took less than a minute to render, so I am happy with that result.

I am thinking of adding caustics. So, if anybody knows:

Is it possible to get caustics inside the fluid (rather than just on the surface)?

How do I use the pre-rendered caustics in the non-PBR render?

Edited March 4, 2010 by Macha

[loudsubs]

I am now working on speeding up the render, without loosing too much of the look. The frame below took less than a minute to render, so I am happy with that result.

I am thinking of adding caustics. So, if anybody knows:

Is it possible to get caustics inside the fluid (rather than just on the surface)?

I'm trying to do the same project, volume fluids falling and colliding in a wine glass. Right now I'm stuck on getting the collision surface right and would appreciate any insight. So far I have a polygon wine glass, I converted it to an SDF volume before bringing into DOPs (using high divisions in both iso offset and DOP static object). When I use ray-intersect mode I get no result, just the same fog from the iso offset, when I choose volume sample, the whole glass gets fill with the red collision geo instead of just the edges of the glass. I can see in your renders that you are getting clean collision geo. Whats the trick to getting the collision geo looking good for fluids?

How do I use the pre-rendered caustics in the non-PBR render?

I think in the vex global illumination shader you can specify under photon maps tab where you want to use your caustic photon maps. And attach vex GI to a light template.

[Macha]

Whats the trick to getting the collision geo looking good for fluids?

It was quite tricky and I never found a really beautiful solution. The problem is that there is always some rough voxelillity (also known as voxellyness or voxelnessity) even at high resolution. In a way there is no antialiasing so the collision is never smooth, unless you have perpendicular walls. So you always get a noisy boundary. To get rid of it I used a smooth sop but the real trick is to smooth only the relevant parts of the fluid, otherwise you loose detail. It is a mixture of distance-to-glass-wall and curvature or normal direction. You need to play around with it as it depends on your setup.