I think that with this it can help to add a modulate_influence with voxel_fog_density. The new attract repulse in the beta does this by default but older versions don’t.
without that modulation the attract affects all space, not just areas with density, and the incompressibility will fight the attraction more.
Duncan Brinsmead
Senior Principal Developer
Regarding the bulging, @jonah.friedman suggested that it might be because the geometry that the attract_repulse uses is offset outwards from your input geometry of the hands. See explanation below.
It is possible to visualize the volume that the attract_repulse node uses to attract the smoke to and by default it uses a detail_size (the size of a voxel) of 0.1 (this detail_size parameter can be set on the attract_repulse node). However as you can see in the image below this causes the volume used by attract_repulse (to the right) to be offset from the input geometry (to the left):
If you set the detail_size on the attract_repulse to a lower value (e.g. 0.01) , you will get a better fit to the input geometry:
You can also adjust an offset inside the attract_repulse node to make the volume seen by attract_repulse smaller than the input geometry:
To set the offset and visualize the volume used by attract_repulse, see the image below. In particular you will need to select the attract_repulse node, right click and select "make editable". Then dive into the node by double clicking. Connect the volume output of convert_to_volume (shown in green below) to the output node's plus sign. And then at the top level in the graph connect this to the output of the Bifrost board.
To set the offset, set the offset parameter (shown in red below) on the convert_to_volume node.
Cheers,
Michael
Here is a scene that roughly does the morph. I tried to create a thing like a hand... the skinny fingers create challenges.
https://www.dropbox.com/s/o0pdgzk1b9the96/aero_morph4.ma?dl=0
https://www.dropbox.com/s/svvwfz4ojwzqxz2/aero_morph4.mp4?dl=0
The graph is as follows:
The mask is set to mask so that there is no effect inside the hand. I made the mask min/max distance slightly negative which pushes the mask inward a little and allows me to use a larger detail size.
The mask controls two effects... one is a subtraction of density each step( the modify_influence). I could have used a dissipate node instead of the modify, but subtracting has a stronger effect on thin density but less of an effect on heavy density... the idea is to slowly eat up the fuzzy fringes. The mask keeps it from eating the density inside the hand. This allows one to be less careful with the amount of density emitted.
The other effect is the attraction, and the mask keeps it from applying attraction inside the hand, which I found tended to cause a slight bias that pulled density out of the fingers. The attraction is along the level set gradient so even inside it pushes inward. The incompressibility of the liquid keeps things from collapsing.
There is a lot of drag on the attract_repulse which allows it to settle faster. I added a little extra drag as a separate influence( not affected by the mask) that then adds drag inside the hand, and keeps the density from swirling around inside it a bit. The motion inside the hand can bring density near the boundary where the modifier tends to kill it.
One needs to emit a slightly smaller volume of density than the volume of the hand because the fluid diffuses over time...especially if there is turbulence or if you have vorticity. However with the modifier that subtracts density one can emit more slop and have it still work. (the mask on dissipation was Michael's idea)
Simulations with a smaller voxel size will have less diffusion and be able to fit inside more cleanly with less reliance on the masked density subtraction.
Duncan Brinsmead
Senior Principal Developer