Originally Posted by Kokosing
I've witnessed fluid sim jobs from several vantage points. From doing them myself to being part of a pipeline where someone else did the sims, to watching a friend's job go into meltdown because the fluid work wasn't meeting the clients expectations. Here are some thoughts.
While rendering fluids is tricky, it's a knowable time frame. But simulation time isn't. I've run sims overnight expecting them to be done in the morning only to find the frame time has quadrupled and I'm nowhere near done. A fluids expert might see this coming, but not always. And sims require lots of refining. As far as I know fluid sims have to be sequential, so there's no way to split the sim over several machines.
There's no such thing as a low-res sim test. Well there is, but it's not going to tell you enough about the sim to be very useful. Good sims need lots of detail and that takes time.
Simulating real fluids is easy. Simulating what your client wants isn't. Anyone can grab a copy of realflow and simulate a glass of water filling up. But what happens when your client asks for the splash to happen in a certain way? Or for a droplet to behave just so? I watched a project go from piece of cake to utter nightmare in a few days because the client was asking for something physically impossible.
Fluids on their own are never enough. A decent bit of fluid work will need a lot of post. You'll need to use the different output files your sim creates. Wet maps, foam maps, meshes, particles, a good fluid comp will take these and produce something magical. Realflow can produce these. But beware, their integration into C4D is not perfect. You will need time to test them. For instance it took me two days to figure out a workflow for using Realflow's wet maps in C4D.
Before you use a fluid sim on a job, think about whether it really has to be a sim. I've jumped into Realflow a few times only to realise a few days layer that I could have done it better another way.
I hope that's useful.
This is all pure gold--thank you Will.
I am usually able to work out a good solution that the client is satisfied with using only particles, mograph tools and sometimes just with deformers. Of course they've all been pretty simple cheats (a liquid pour with animated deformers, a splashing eye drop, running rivers and streams, things like this). I think it is always good practice to see if you can pull something off quickly and effectively with the tools we have before resorting to sims.
Your comments about getting a sim looking good right off the bat, only to have a client ask for specific subjective changes is a scenario I know well--as do many of us I'm sure. This also goes for any kind of GI or physical rendering approach, particle work, etc when under a tight deadline.
And this is one reason I try to steer clear of sims, or anything that is not tried and true where I have complete control and know I can get it to render quickly. I will definitely keep these suggestions in mind, as should we all.