Final simulation and breakdown for the shots.
Sim time: 54:00:28
So since the last post a lot has happened. We finished all the visual effects for the movie and everything is now ready for audio and grading. Meanwhile me and Ivaylo created a video on how we got to finish our most difficult VFX shot. The video goes very in depth from where i did the simulations and how it ended in compositing. The video has already been online for a while and we got a nice amount of views as well. Also being featured on some VFX sites helped us gain some popularity.
With the last iteration I increased the size of the pool to get less of the spilling, although it happens now it is easier to mesh at the end. Since this sim is going to be two shots I had to resim it with a different ship model and more frames. Posting more results soon.
Sim time: 20:28:45
Finally got really good interaction with the particles in this shot. I had so much trouble getting the water to punch up like it does in this scene. Apparently domain offset was the issue here, I tried before at 3 meters but had to set it at 7.5 (meters!) which is 750 times larger than the default. On the other hand it also makes sense since the robot is 11 meters high in “real” life.
Probably going to do one more iteration.
Recently I have been busy completing scene 16 and it has been a pain. I was not getting the simulation under control.
The problem with the sim was that I did not get the interaction with the robot down as I wanted it to.
As you can see the robot is hardly interacting with the robot. Most of the time when this happens you check how the isosurface is. Inside realflow it was looking pretty good, so I tried adjusting the interaction factor on the robot. This actually did create the splash that I wanted but the physics where all over the place.
The plashes for example had a high velocity and kept on rising way to high. Not the result I was going for.
I fiddled with domain and surface offset on the robot and tried various combinations but to no avail. I’ve ran several simulations where results were exploding hybrido tanks.
Isosurface was eating particles
The next thing I tried was increasing hybrido substeps. This also did not help the fluid. I was still getting the same results.
I though maybe the resolution has something to do with it and since I had a weekend of time I decided to go crazy with the global cell size. I put it down to .3 and the sim contained 90 million particles and generated 500gb of data through the weekend but again, the sim was not looking good.
Back in May we had the FMX festival and I had the chance to meet Gus and Luis from Realflow, I showed them some robot shots done with Realflow and asked me if I would like to be in the beta program. Since I tried Realflow 2015 alpha I decided to ask questions on this issue on the beta forums.
I posted an extended topic on the hybrido interaction and how I wanted it to look. RealFlow Product Manager Gus gave me some tips about scene optimalization and tips on the robot isosurface.
He told me to combine the robot into one single mesh and import it either as a multibody or in a SD file. Since I had trouble in the past trying to create a proxy model for the robot I had to check if there was another way possible.
The problem was that after using mesh -> combine in Maya the robot lost it’s animation. However baking the animation first and then using mesh -> combine worked. Now I had the entire robot as one single object. Before this I cleaned up the rig and reduced the poly count significantly and got rid of a lot of the smaller objects. Now I have a perfect proxy model.
I used advice from user pixellusion to introduce some randomness in the hybrido scene and added an ocean daemon. I decided to leave the interaction factor at 1 and tried other values.
I also double checked FPS, scale and such but these where all good. Drag force was at default.
I’ve also used Gus’s tip on getting more object friction in there but it didn’t create the splash.
I’ve tried fiddling with Core/Surface particle sampling but to no avail. It did not create the splash that I wanted. I however found an solution for this whole problem.
I am currently using 4 surface samples for the surface and 2 core samples for my sim. Also i’m using a 3 meter surface bandwith since our robot is 11 meters in “real” height. (Your tip was used here pixellusion, thanks).
Eventually Luis had the right answer. It was the domain offset. I’ve tried before with a domain offset of 3 but it didn’t work. Since the default is 0.1 I already thought i went over the top. However last friday I tried a value of 5 and it started interacting the way I wanted.
Now after some fine tuning after friday I’m getting the result I wanted with a value of 7.5 as domain offset (that’s 7 and a half meter!). Considering the robot size this might be logical, I’m not sure to be honest :D.
I’m still using a relatively low depth water but the splash is working perfectly so far. I am still simulating as I am typing this. Here is the half-way result, VERY satisfied so far!!!
Look at the bulge that’s happening at the bottom of the splash, the interaction is really good.. Exactly as I wanted.
Also the speed is very acceptable. 15 million particles in this sim.