We are workign hard on our furry charracter fully build with Ox & Lucid and it's time to give some feedbacks.
Here's a preview of the animation we are trying to simulate, a sort of karate AKA :
And a close up of an arm witch could be a real node bag
Ornatrix 4.5.0.7782 & Lucid 1.2.0.7779
We are getting closer to something 'not bad', but for me, it's still far from being clean & 'cinema quality'.
Here's some feedbacks from my FX artist.
- Could be nice to unblock some parameters for animation, like Damping or Dissipation (lucid flex settings), to adapt simulation to the style of animation.
For the moment, the only way we found is to blend with a morpher between 2 simulations. It's pretty hardcore to maintain.
- Possibility to simulate a part of the fur/guides and not all.
- Have a adaptation of lucid particules in size and number along guide lenght. For the moment a 30cm long guide (arm) has the same amount of particules as a 2cm one (body / arm). So we build Lucid to make long guides being nicely sim, but the short ones are going crazy (because all lucid particules are one above the other).
Possibility to have a differents size along lenght could also be interesting.
- Simulation is pretty unstable on very slow animated mesh. We have tryed to stresh a point_cache of a fast animation, to avoid increasing too munch subdiv. But slow part are now very slow and looks like the simulation is shaking.
- (Ornatrix) From my side, i'm working on a solution to bake a simulated animation and/or loading a Fur on a huge amount of charracter at same time. (furry crowd)
I've tryed the Ox_alembic export https://vimeo.com/166275457
But the files size are going crazy ! +500Mo for 50 frames of a teapot with 100k hairs. How is going to be a +1.5M hair furry charracter ? ^^'
Exporting only guides could be a possible solution ?
- I'm looking for a more robust fallback for hair dynamics. Do i have any other solution than simulating over Maya ?
Thank you for the feedback, it is very valuable! Glad to see the current results, I think it looks great with some of the issues you're having.
> Could be nice to unblock some parameters for animation, like Damping or Dissipation (lucid flex settings), to adapt simulation to the style of animation.
Should be available in next nightly build
> Possibility to simulate a part of the fur/guides and not all.
We will add this in one of upcoming versions, I just logged it. I think the best way is by using strand groups (like curling/length/etc modifiers in Ornatrix)
> Have a adaptation of lucid particules in size and number along guide lenght.
Just logged this, we will work on it.
> Simulation is pretty unstable on very slow animated mesh.
I see the problem in the video. We recently made some changes/fixes to hair simulation, perhaps you can try it again in the latest nightly? Otherwise, if you can send me some basic scene showing the problem it would be very helpful to track it down.
> Simulation baking
Have you tried applying a point cache modifier onto guides and then storing the animation into a pc file? You can then load it and animate the guides. Alternatively there is Ox Strand Animation modifier where you can save into .oxh files.
> I'm looking for a more robust fallback for hair dynamics. Do i have any other solution than simulating over Maya ?
Aside from Ox Dynamics (which uses PhysX) and/or using cloth animation with either Lucid or 3dsmax cloth modifier I don't know of any solution. I think we can improve Lucid hair simulation to a point where it will solve your current problems.
- Concerning ground stand : my groom artist is not using it at all, he preferes to work with maps.
- One another boring bug : whenever you change a value in the Ox stack of the character, we have to regenerate (manually) the clusters or it will bug after frame 1 (see video 72)
- We are getting close to a cool simulation now. I hope that the last part of cleaning unstable areas will not be too munch time consuming.
The very short strands have a minimum number of particles which is needed for simulation to be stable. I can try to make this number smaller or maybe add an option to make very short strands not part of simulation?
I am not sure I understand the problem with clustering. When you generate the clusters hair indices get assigned to each cluster so you must be careful not to change the topology afterwards or you may need to generate the clusters again. It is also important to generate the clusters when the character is in a "neutral" pose
