Carrara Dynamic Hair

DartanbeckDartanbeck Posts: 21,106
edited December 1969 in Carrara Discussion

I have a hair do that I made in Carrara that I really like.
The problem I've been having is in animating it.
I've used a proxy figure to help with the vertex counts.
The problem seems to belong to the shader. When I remove the frizz, it seems to animate smoothly.
But with the frizz applied, it looks a bit 'electric' as it spazzes from one frame to the next.

Any ideas?
I wonder if it's possible to keyframe drapes at various points - rather than using "Simulate Hair"?
Or perhaps I'll try converting to another modeller and making it conforming?

«13

Comments

  • wetcircuitwetcircuit Posts: 0
    edited December 1969

    Are you using Clumping at all?

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    No go on the idea of keyframing different drapes
    Coverting to vertex model leaves you with nothing - just gone.
    Probably hair is more of a special effect than anything else.

    Are you using Clumping at all?Yeah.
    Hair Clump Shader
    Count + 40%
    Clump = 60% Variation = 34%
    Rate = 0%
    Radius = 0.10 Variation = 47%
    Clump in Local Space = unchecked (maybe I should check this?)

    Hair Kink Shader


    Root = 1%
    Tip = 40%
    Frequency = 92%
    Frequency Scaling = 504%

    I don't have a Frizz or Wave shader.
  • DartanbeckDartanbeck Posts: 21,106
    edited January 2013

    Ideally, I'd like the hair to not move around much at all. Just follow her movements. Unless, of course, she's running, fighting, standing in the wind, etc.,

    Do you animate with Dynamic hair much, Holly?

    Post edited by Dartanbeck on
  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    The rest of my settings:

    Hair Thickness Shader
    Root Thickness = 50%
    Tip Thickness = 9%
    Hair Length Shader
    Length = 100%
    Length Variation = 10%
    Hair General Tab
    Length = 2.4 ft Thickness = Very Large Scaling - 1.00
    Render Quality 500% Placent Type: Geometry Based
    Segments per Hair = 24 Hair Count = 4000
    No Reverse Normals (unchecked)
    Mode = Long Hairs Override Segment Count = Unchecked
    Segments per Guide Hair = 12
    Guide Count (greyed out) = 49
    Guide Sampling = Custom
    Symmetrical Sampling = unchecked Auto Grouping Enabled
    Distance = 833.33 ft
    Angle = 60.2 degrees
    Distance Along 5.0
    (It was a while ago that I made this hair - not sure where these figures came from...)

    Hair Dynamics Tab
    Stiffness
    Root = 100%
    Hair = 5%
    Shape = 5%

    Damping Air = 50%

    Density = 2.00

    Use Scene Time Range

    Quality
    Simulation = 26% (Trying to get it to make fewer calculations)
    Hair = 100%

    Collision Enabled
    Distance = 1.0

    Draping Iterations = 100

    No Applied Forces at this time except for a simple head turn during a walk cycle

  • evilproducerevilproducer Posts: 9,040
    edited December 1969

    Maybe a stiffer shape would help? I've tried a couple hair animations, but my system is getting long in the tooth and I don't have the patience.

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    Tried that :down:

  • FenricFenric Posts: 351
    edited December 1969

    I tried a number of experiments with just a ball... it doesn't look like "frizz" is animation-friendly: whatever random function they use is a bit too random. I was able to produce several variations on ugly, but nothing useful.

  • WendyLuvsCatzWendyLuvsCatz Posts: 37,712
    edited December 1969

    I have LAMH for DS4 which can create geometry
    this can be conformed to Genesis using transfer tool saved as an asset .duf file and will import as a Genesis hair item or imported into Carrara as an obj selected along with the hip of any figure and in animation tab attach skeleton, weightpaint if needed.
    not as good as Carrara hair, no collision for styling and brushing and no draping but that might be coming soon.
    I myself have been using it to create furry clothes for Genesis figures in iClone but it is very highpoly.

  • wetcircuitwetcircuit Posts: 0
    edited December 1969

    If you use clumping and want to animate you MUST check local space...


    Are you using Clumping at all?Yeah.
    Hair Clump Shader

    Count + 40%
    Clump = 60% Variation = 34%
    Rate = 0%
    Radius = 0.10 Variation = 47%
    Clump in Local Space = unchecked (maybe I should check this?)


  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    I'll try some more...
    Thanks for your help.
    I really like this hair - and am hoping to get it to work.

  • 3DLust3DLust Posts: 230
    edited January 2013

    Hi Dartanbeck - I really hope you get the hair to work.

    A couple of tips and tricks that seem to work for me:

    * Increase gravity in the Physics settings for your scene to -500.00

    * Only use carrara spheres for the collision objects - make sure the hair is set to
    collide with nothing except the spheres. If the hair is made on a cap, make sure the
    hair is set to not collide with the cap.

    * In Hair Dynamics, try root stiffness 80%, Hair 10%, Shape 5%, Air Damping 20%, Density 1.0 -
    If you want it to follow her body for simple movements, try increasing the Air Damping to 75%.

    * For Segments per hair, you can use a high number like 50, but then check Override Segment count
    and set at something like 12 Segments per Guide. The simulation will run pretty fast, and you can bump
    up your simulation quality back to 100%.

    I was able to get a decent simulation with these settings using Kink and Clumping in the shader, so
    I hope it will work for your long hair.


    I hope this helps!

    Here is a quick video that demonstrates the difference between Air Damping of 20% and 75%.

    http://youtu.be/KjQY-8f1Efo

    Post edited by 3DLust on
  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    Wonderful input, indeed! Thanks Man!

  • 3DLust3DLust Posts: 230
    edited January 2013

    No problem : )

    I'm doing a test with some longer hair, and you may need to bump the Segments per Guide up to
    a larger number. When I increased the length of the hair while set at 12 segments, it started some
    vibrations, but if I changed it to 24 segments, the vibrations stopped.

    Post edited by 3DLust on
  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    3DLust said:
    No problem : )

    I'm doing a test with some longer hair, and you may need to bump the Segments per Guide up to
    a larger number. When I increased the length of the hair while set at 12 segments, it started some
    vibrations, but if I changed it to 24 segments, the vibrations stopped.

    Cool info. Thanks a bazillion!
    Keep this up and I'll buy all of your products again!
  • 3DLust3DLust Posts: 230
    edited December 1969

    No need to buy them again - hehe - I just love seeing the renders you do with them. :)

  • DartanbeckDartanbeck Posts: 21,106
    edited January 2013

    Okay.
    Following your advice, without making any other changes, it turned out much better. I forgot that I had it set with a hair count of only 4,000, which may be why I set it at Very Large for thickness, which I actually kind of like. But it does get a bit Rag-Doll looking close up.
    I really like this hair. Just knowing that others are getting good results was enough for me to continue on and make it work. It's been sitting in my browser all Summer, perhaps longer. I just thought it wouldn't work.

    For stills, I think it's the best hair (Carrara hair, in general) I've ever seen. I wish I could get away with making a conforming hair that worked like this - instead of being so heavy with alpha maps. However, I think this may take just as long to render. When I watch it render a still, it seems to be faster by a margin. Animations seem to be where it really chews time.

    Oh... and I turned the clump shader local, silly me for not doing that in the first place. Now I'm going to go for more hairs, thinner hairs, a touch less air friction, and possibly a touch more downward gravity. I can already tell that if I use this hair, I'll have to animate in sequenced image files - as it has the tendency to drop the roots for a frame. Not sure why, but who cares. Also, instead of using Carrara spheres, I'm using the V4 proxy conforming figure that comes with PhilW's Advanced Carrara techniques dvd set. If I can't get this to settle down to an acceptable level (your shorter Kinky Hair test on a globe looks perfect to me), I'll try converting the proxy to the primitive modeller and see if that makes a difference - other than that, i'll try a pile of spheres parented to the head, chest and shoulders at the very least. Tougher is that when your going for this long of hair. She found out that I love long hair. So, without telling me why, she didn't ever get her hair cut for our first eight years of marriage! lol
    I never complained!

    Dyn-Hair-Test2.jpg
    720 x 1280 - 199K
    Post edited by Dartanbeck on
  • wetcircuitwetcircuit Posts: 0
    edited December 1969

    GLAMAZON!

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    GLAMAZON!
    What a perfect name for this hair! Thanks Holly ;)
  • FenricFenric Posts: 351
    edited December 1969

    Really cool. Glad to see you got it working.

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    It got a little better. I'm noticing, however, that everybody else, who's getting good results, are using a much much higher hair count - so that's my next step. Its getting so much better that I may just have to plan on the switch, from here out - a new Rosie. It's been a while... maybe I'll work a little more on all of her. Get her ready for production.

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    Much higher hair count, slightly more gravity, and 50% Air friction.
    I can see a big difference in appearance, we'll see how well it animated when the render is done

    Dyn-Hair-Test3.jpg
    491 x 615 - 51K
  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    I forgot to turn a bunch of stuff back on for the above render. I like the texture of the hair itself much better in the previous image at the top of the page. We'll see how it looks tomorrow.

  • 3DLust3DLust Posts: 230
    edited December 1969

    Much higher hair count, slightly more gravity, and 50% Air friction.
    I can see a big difference in appearance, we'll see how well it animated when the render is done

    I really like this one - the hair looks much more real now - nice work!

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    Have to give up for now.
    Can't get the twitching to stop during animation renders no matter what I try.
    Maybe next time I can waste a week waiting for renders I'll try again using different shader settings altogether. The bummer with that is that it took forever to get just that right look that I wanted. Too bad.
    It animated great before, when I took away the curls.

  • thoromyrthoromyr Posts: 452
    edited December 1969

    Fenric said:
    I tried a number of experiments with just a ball... it doesn't look like "frizz" is animation-friendly: whatever random function they use is a bit too random. I was able to produce several variations on ugly, but nothing useful.

    Frizz is just unfriendly. I need to update my notes on hair to reflect this, but frizz -- even without animation -- very quickly goes into some weird state where the hair (or some strands of it) are scaled to extreme size (except for diameter). The result is a hash of apparently random lines across the scene with all segments at significant angles.

    I haven't done enough cross checking to be sure of the whys and wherefores, but I suspect it is an issue with too many segments (too many strands or just too high of a segment count).

  • evilproducerevilproducer Posts: 9,040
    edited December 1969

    thoromyr said:
    Fenric said:
    I tried a number of experiments with just a ball... it doesn't look like "frizz" is animation-friendly: whatever random function they use is a bit too random. I was able to produce several variations on ugly, but nothing useful.

    Frizz is just unfriendly. I need to update my notes on hair to reflect this, but frizz -- even without animation -- very quickly goes into some weird state where the hair (or some strands of it) are scaled to extreme size (except for diameter). The result is a hash of apparently random lines across the scene with all segments at significant angles.

    I haven't done enough cross checking to be sure of the whys and wherefores, but I suspect it is an issue with too many segments (too many strands or just too high of a segment count).


    I've had this issue as well. Come to think of it, it may have been with a high hair/guide hair segment count.

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    Trying something different.
    As I have come to expect that it is the shader itself that is causing the jitter, or perhaps the combination of that and that the hair is so long, I am going to try this method.
    I am currently using PhilW's Proxy Figure. Not the one that you conform to V4, but a stand-alone proxy model that you animate. I'm leaving the scene empty of everything for now, put a glowing green shader on the proxy model, and am filming the animation with an alpha channel into a sequence of Targa images.

    My hero character has a different scale than the proxy. So if this works okay, I'll have to make my own poxy, or just edit this one. If it doesn't work out okay, I'll try the same effect, but with primitive spheres in place of the proxy figure.

    Howler accepts sequenced Targa as either an animated brush, or a green screen (or many other composite types) in the swap channel.

    So I intend to remove the worst jitters in the hair animation by deleting frames evenly throughout the animation sequence, and then use the motion prediction module to tween differences between the remaining frames to get the sequence back to the original length, which I'm hoping will both remove the sporadic behavior, as well as keep the full motion that the physics engine is providing.

    Copy that sequence over to Swap and add a new sequence of the hairless hero, which uses the same camera, and is scaled to fit the hair placement. Then combine with swap using green screen.

    If this works as easy as I just made it sound, I will actually save render time by doing it this way but also get the added benefit of the beauty of Carrara's dynamic hair.

    GreenScreenTest1a.jpg
    391 x 615 - 165K
  • DartanbeckDartanbeck Posts: 21,106
    edited April 2014

    As you can see, this test is being done with a very low fidelity hair of only 7,000 hairs. I just need to know if the method will work. If it does, the hair will get made much nicer, and I'll put all of the advice given in this thread to the simulation.

    EDIT: Also, I don't recall if the alpha can be respected in Swap. If not, the brush keyframer will, which may be the best route for this anyways, as it gives full visual control and the ability to make keyframed alterations.

    Thanks everyone!

    Post edited by Dartanbeck on
  • evilproducerevilproducer Posts: 9,040
    edited December 1969

    As you can see, this test is being done with a very low fidelity hair of only 7,000 hairs. I just need to know if the method will work. If it does, the hair will get made much nicer, and I'll put all of the advice given in this thread to the simulation.

    EDIT: Also, I don't recall if the alpha can be respected in Swap. If not, the brush keyframer will, which may be the best route for this anyways, as it gives full visual control and the ability to make keyframed alterations.

    Thanks everyone!

    Maybe a shadow catcher instead of green screen and rendered with an alpha channel should also work. May prevent some color spill between strands and it should act as a mask for those hairs behind the figure.

  • DartanbeckDartanbeck Posts: 21,106
    edited December 1969

    I was wondering about that... so that might just make the area where the shadow catch is as alpha?

Sign In or Register to comment.