Maya 6 Hair Problems

[prettyh8machine]

I've just started testing the Maya 6 Hair and I'm having a few problems:

• if I try to select certain CVs with marquee select tool, it takes ages to select them. also, if some or all start cuves are selected, view port navigation gets unusually slow.
• at times when the start cuves are falling down and so are the rest curves and there is no gravity or other dynamic forces, playback results in hair tryiing to move up for a while and then very slowly they start falling down towards the start cuves but they just dont get relaxed and kinda oscillate. funny thing is that same settings worked with the tutorial shipped with maya and they almost always work except these moments. I just wanted to know that is there something that i'm missing or what?
• can we apply normal dynamic fields (like newton, air) on certain hair curves? because i really feel it hard to "comb" the different portions of hair by using a dynamic field on all the hair.
• I also tried constraints but even if the lock length is on for the selected start curves, constraints like transform and stick, if moved here and there will result in hair getting stretched like strings. what's with that?
• For rendering them in Renderman, the only way i've found out is to first rebuild the output curves to degree 3 curves and then render, otherwise mtor issues an error. what's with that?

Please hlep.

cheers

[duncanbrinsmead]

The selection draw speed problem is related to the follicle draw. If you hide the follicles or disable their display for your view then the refresh should be significantly faster when you have many items selected( it doesn't really matter if the many items are hair curves, cvs or unrelated geometry).

If the start curves are not initially in the rest position( either a straight light or the shaped defined by a rest curve ) then when there are no forces acting they will move towards the rest shape. Rest curves sometimes result in positions that are not stable and do not settle over time ( curling shapes are particularily unstable and it is good to handle curls with the hairSystem curl attributes rather than by using rest curves ). Increasing the damp value a little will help most cases. In extreme cases you could use a very high level of damp just on the problem follicle(s).

The fields are applied on a per hairsystem basis, although you can split hairs into new hairsystems by assigning a new hairSystem. The different systems can then have different fields. Afterwards one could recombine the hairSystems by assigning the hairs back to the original hairSystem.

One thing, however, that might be easier is to use a transform constraint and translate/scale it during interactive playback( then set start from current and delete the constraint ).

The lockLength attribute only affects the curves they are applied to. The start curves are simply a starting set of points for the simulation. When computing the dynamic simulation we need to iterate towards a solution. Computing zero stretch with the method we use would require an infinite number of iterations. The constraints stretch the hair because the stiffness of the constraint is generally much stronger than the internal stiffness within the hair. You can lower the constraint stiffness value if you don't want it to be able to stretch the hair.

For rendering in renderman, you might look at the pfxStrokes command. This outputs all the paint effects hair( including curling effects, haircolor and widths) to
an ascii file.

Another option, although it is more memory intensive, is to again use the pfxHair output from the hairSystem but to convert pfxToPoly (one can also use colorpervertex if desired ) This creates a mesh of triangular shaped tubes.

Also if you need degree 3 output curves, rather than rebuild them it would be more efficient to set the output curve degree on the follicle nodes to 3 (instead of the
default of 2). (I'm not sure why you get that MTOR error)

Duncan

[prettyh8machine]

Hey, thanks a lot for the help. You almost answered all my queries satisfactorily except that in the case of rest curves not being in a stable position, I dont really think that's the problem. At least in my case, I even tried creating hair of length = 1 and then let them fall down with the default gravity. Once they were just above the scalp, i set the start and rest curve both to the current position. Now when I set the gravity to 0 and playback, the hair still try to move up at first and then fall down but never come close to the rest curve. I tried all values of damp and while they do affect their movement, they still go up instead of stayin there oscillating cuz i had set the stiffness = 1 as well. One thing that might have caused problem was that I was also using sphere collision constraint with sphere being almost under the head, definitely below the rest curve positions.

I read in one of your posts that rest curves are generally supposed to be a little above the start curve. While it makes sense, if we set both to the same position, shouldnt stiffness=1 and no forces make the hair stay where they are from the start till the end??

One more thing. If I want to create the most normal haircut with smaller size hair covering the temples and back of the head and larger size on top, what's the best solution:
- make one hair system and then use hair scale tool to scale the selected follice, or
- create a smaller hair layer first and then create other medium sized and larger sized hair layers later

and oh, is it more feasible to add a lotta follicles on the scalp like in real life or less follicles with higher number of hair on each follicle? I think definitely latter but can you mention a scenario where former might be feasible?

Thanks a lot for the help man, I'll keep posting as I keep coming with more queries.

Thanks again.

[duncanbrinsmead]

"I read in one of your posts that rest curves are generally supposed to be a little above the start curve. While it makes sense, if we set both to the same position, shouldnt stiffness=1 and no forces make the hair stay where they are from the start till the end?? "

This is correct. I double checked it and this is the behavior for me. Create hair on a sphere.. play the
simulation a bit.. set start and rest from current. Make gravity zero on the hair system.. go back to the start frame and play. The hair should not move(providing the underlying surface does not move).

Perhaps your collide sphere is pushing the follicles a bit( check the clump width and collide width offset ).

There are advantages and disadvantages of using multiple hair systems:
-easy to edit and modify the separate groupings
-constraints do not work across multiple hair systems,
so you would need to create more. It is not possible to
do hair-hair constraints between separate hair systems.
-it is more work(but not hard) to keep some attributes like hairColor in sync.
-hair self collisions do not work between different hair systems(it is better to use constraints when possible, however)

I can't really say what will work best for you. Try experimenting with different approaches. You mght find
that some layers.. like an outer frizz.. are best handled with separate hair systems, because this will make editing easier.. you can also hide some layers for faster feedback when working.

Duncan

[prettyh8machine]

Well yes you're absolutely right; even if the start curves and the rest curves are both set to the current position resting almost on the scalp, in the absence of any dynamics, the collision sphere somehow pushes the hair to move up with a jerk and then gradually they come down to the rest curves and most of the times, they just dont get relaxed. This is very annoying and even if the stiffness of the hairConstraint is modified, the hair do get pushed up a great deal. Is this what's supposed to happen or there's something wrong with it?

Secondly, I dont see any randomize option in the hair paint tool which makes the painted portion of the head having too symmetrical the hair follicles. Is there any workaround for this or there is an option somewhere hiddden?

Thanks.

[duncanbrinsmead]

You should make sure that the collision sphere does not overlap with any hairs(including the clump widths), especially towards the hair base. Note that you may need to increase damping(or drag) to get hair to settle. Technically if you give an undamped spring a push it will wobble forever. However there is an additional problem with rest curves not always conserving energy, in which case one sometimes needs to use very heavy damping.

We were going to add a randomize option on the paint but it did not make it in. Here is a simple script that you can use to randomize the positions of follicles after you create them( it simply randomly offsets the parameterUV values on the follicles). Put this in a file called randomizeFollicle.mel in your scripts directory. Then select the follicles to randomize and type:
randomizeFollicles
Make the value for range as small as will work for you. Typical values would be 0.02 or so.

Duncan

//-----------------------------------------------------------

global proc randomizeFollicles( float $range )
{
string $follicles[] = ls -sl -dag -type follicle;
int $numSelectedFollicles = size($follicles);
if( $numSelectedFollicles > 0 ){
string $follicle;
for( $follicle in $follicles ){
float $u = getAttr( $follicle + ".parameterU" );
float $v = getAttr( $follicle + ".parameterV" );
$u += rand( $range ) - ($range*0.5);
$v += rand( $range ) - ($range*0.5);
setAttr ($follicle + ".parameterU") $u;
setAttr ($follicle + ".parameterV") $v;
}
} else {
warning( "No follicles selected to randomize." );
}
}

[prettyh8machine]

Thanks for help Duncan, but you said:

"You should make sure that the collision sphere does not overlap with any hairs(including the clump widths), especially towards the hair base"

Now if you mean that I should place the sphere inside the head such that it's below the base of the hair (i.e. even below the follicles), then that causes the hair to penetrate the scalp and create another problem. Damping or no damping, current position hair do get a horrible push upwards when it's played back as long as there's a collision sphere. It minimized by placing the sphere well inside the head and tweakin with the stiffness of the hairConstraint etc. but still, it is a problem.

Btw, I havnt looked at it really but Im hoping I can edit the paint hair tool UI script myself and incorporate the randomization script provided by you, hopefully, can I?

And oh, I almost forgot...sometimes it gets really hard to fit a collision sphere that correctly cover the whole head and ears as well, we can only approximate. This almost always work except when u try to make very unstiff hair that almost are sticking to the scalp cuz then they get penetrated. Is there anyway to make that sphere deform in a more flexible way or there's some other way out?

Thanks

[duncanbrinsmead]

The collision sphere will repel all hair cvs from the sphere, including the start cv. If it collides with the base of the hairs then when you start the simulation these bases will jump to the outside of the sphere and cause a movement down the hair. You could use the collide width offset or clump width to scale the effective width of the hair.. the first cv width is always zero if pointlocked to the scalp, so clump width alone should not make it jump as long as the sphere does not overlap the follicle.

To better approximate the shape of the head you may wish to use several collide constraints. In sensitive areas like the forehead and cheeks you may wish the
collide constraints to protrude beyond the surface somewhat, as long as it does not overlap follicles.

If needed you can combine geometry collisions with the collide constraints, but this is fussier to deal with and I would generally avoid it.

You can edit the paint script to add random offsets, but of course you should be careful not to break things... also you will need to update your edits when you get new versions of Maya. You will need to be fairly comfortable with mel. The file is:
hairCurvePaint.mel
Go to the routine setHairPaintValue and search for the variable uv. This variable controls the position of the
newly created follicles.

Duncan

[prettyh8machine]

Hey Duncan, thanks for the help.

Regarding collision sphere, you mentioned the situation where the collision sphere is either overlapping the base of the hair or the follicle. In my case, I had the collision sphere well under the head which means that there was a considerable distance between the follicles all over the scalp and the collision sphere...and even then, the hair would get pushed up. In any case, I'll double check it and probably post a wireframe image as well to make it further clear. I think there must be some part of the hair curve touching the sphere or something like that. Let me just see.

Thanks for giving me the multiple constraints idea, I can't believe I couldn't think of it myself

Also, thanks for pointing me to the exact location where I'd be needing to make a change. Dont' worry, I'm fairly comfortable with MEL and hopefully I'll figure it out. I'll let you know in any case.

Thank you.

[prettyh8machine]

Hey, now since I've known lots of methods and have done experimentation with styling the hair and modifying their looks, I've now started working on the animation.

Unfortunately, the hair just dont adhere to the rest curves at all and instead just fly all over the place. It's not like they get detached or broken, but they move very slowly as compared to the head.

I've been experimenting with the female head model shipped as a tutorial scene with Maya 6.0. I've set a parent constraint b/w 'locator1' and the sphere collision constraint. Now when playing the simulation, I rotate the locator1 and move the head here and there quite naturally, the hair are flying as if they're in water or something. Very slowly and each clump comes back to the start curve so gradually as if no gravity or force exists. I've tried almost all settings of stiffness, stiffness scale ramp, drag, damp, gravity, iterations etc. They might look ok if I have to simulate long silky light hair but if I'm trying to simulate mid-sized hard hair, kinda stuck to the scalp always, do I have to keep them constrained with constraints all the time or what? Shouldn't they just stick to the rest curves as much as possible, specially if the stiffness is 1?

I hope it's a pretty basic problem I'm having and there must be an easy solution.

Thanks in advance,
Cheers

[prettyh8machine]

Ok, now i've literally tried all the possible values of all the attributes that could possibly affect the dynamics of the hair. They just dont move right. They even stretch a long distance and then comes back to the scalp suddenly. I've still not used any constraints so far but at least without them, they're lagging way behind the parent. I increased dampness to very high levels but everything seems useless.

Please help.
Cheers

[duncanbrinsmead]

It sounds like damp and drag are perhaps too high. Also you need to increase the iterations( and make lengthflex zero) to make the hair stiffer and less stretchy. If you have very fast moving geometry you might need values like 100 for iterations. Note that to keep the hair flexible, yet not stretchy, you need a high iteration level, but may also wish to lower the stiffness to compensate. A transform constraint(parented to the head with Uparam = ~0.2) will help you lock the hair onto the head as it moves. Much of the locking in normal hair is self collision, which is generally not practical and also requires hair/hair friction which is not yet supported.

Duncan

[prettyh8machine]

Hey, I was doing everything except that I didnt raise the iterations level that much...so now even though the simulation runs a little slow, at least the hair are moving in my desired way. Thanks for the tips.

Thankyou

[duncanbrinsmead]

Note that if you need a high iterations count, perhaps you have more segments in the hair than you need. In some cases it is better to rely on the subSegments to get smooth detail(subSegments are render only segments), and simulate on hair with fewer segments.

Duncan

[prettyh8machine]

ok i'm having a new problem:
the head animation i'm using is pretty fast (around 270 degree rotation in ~25 frames) and the hair, despite using damp=100, iterations=100, static clung=1, high gravity and everything, still move silkily. while it might be possible that this should behave as it does, in reality as well, what other parameters i can change to make it move more slowly? any other workaround?

Thanks

[prettyh8machine]

ok nevermind. i was askin the wrong question. i actually wanted to hair not to fly so high and i managed to control that using the wonderful transform constraint with uparameter=0.2. cool.

Just another question:
while the best way of makin a pony is to make a single hair clump with a high number of hair, this causes the pony to have all hair stuck together always (naturally cuz they're strokes, not actual hair). that's fine as long as pony isnt moving too vigorously but when it is, it doesnt look good at all. one way could be to add extra clumps with low hair count (may be 5-10) just with the main pony clump to simulate the hair in the pony tail moving apart but I was just wondering if there's another workaround built-in or something like that. Is there any?

Please help.
Thanks

[duncanbrinsmead]

One possibility might be to create a single(non-rendering) hair for the motion of the ponytail, and then attach one or more constraints to this hair that constrain the main hairSystem hairs(not the single hair).

To attach the transform constraints to the single hair you could do the following(which will preserve the correct twist values ):
*create the single hair as a new hair system with pfx output.
*select the hair then do pfxToPoly( a thin tube results )
*select the pfx to poly mesh and create a new hairSystem on it with a small number of hairs(or just 1) to be used for the constraints. This is just to get the follicle transforms. You can delete the new hairSystem node as well as the input start position curves.
*presumably you originally had a complete hairsystem with the hairs pulled back into a pony tail shape. Now select the hairs to be constrained into the ponytail and create a transform constraint. Parent the transform constraint to the follicles generated in the previous step.

Hopefully that describes the process enough to get you started.

Duncan