melvelvin
patreon
WIP#18: sticky!
🕑 Added 2022-03-04 15:52:32 +0000 UTCHey y'all! I've finally got around to doing some more serious experimentation with methods of depicting the nut. Recently people have been experimenting with geo nodes and I have to agree - the results are really promising! using particles and remeshing can produce results far more controllable and real-ish-tic looking than a fluid simulation, in a fraction of the time. Under controlled circumstances, they can even pool, gloop, drip and dribble in ways that it would be prohibitively difficult to do manually! Very exciting.
This was kind of a stress-test to figure out the limitations of the technique and figure out ways it could be improved. In no particular order here are my conclusions:
- dynamic paint can be used to enhance the effect, but can also very easily look wrong. Careful tuning is needed
- As usual with particle sims, the scale is all wrong. The liquid flies through the air far too slowly.
- The simulation is prone to globbing up in an unrealistic way, and this behaviour seems very difficult to tune out. This causes three problems: First, while the stringing behaviour is far better than previous methods, for "sticky ropes" to look right, they need to be smooth, this method still produces very rough results; Second, some very small droplets can still be created, these are very unrealistic; Third, gigantic rolling balls of the fluid can form as many particles agglomerate into a single mass, ideally it should be doing almost the opposite
- The particles can still be sloughed off the skin by sharp movements, giving is a "brittle" or non-newtonian feel. This can partially be avoided in animation, but ideally the stickiness should be improved.
- volume preservation is a joke with this method. (it's a hard problem at the best of times with any method) Some (more?) repulsion between the particles might solve this?
- Because of the use of geometry nodes, there is an opportunity to "mix in" manually created or animated geometry to enforce more wanted behaviour, like stringing and smearing. This may solve a two or three of the above issues.
- So far, the workflow is more than acceptable in terms of weight - simulation times are short, iterations are low and work time is manageble. Any improvements will need to keep it that way.
