I will beb donig a project about character animation. I am think about whether I will choose Houdini or Maya. The thing that I am concerned is “Hair”. Well, I am a maya user but I am a beginner in Houdini. I just want to know how people deal with Hair (dynamic) in Houdini. Can it achieve Maya 6's hair quality in terms of its dynamic simulation?

From my poin of view:

As far as exist the hair solutions for houdini remain as inhouse developments in the studios - little bit slow, little bit buggy, little bit restrictive worklflow.

Seems that the Hair stuff in Maya is very good solution - kick-but hair-to-hair collisions, constraints, hair-to-geometry collisions, rest curves ( goals -startPosition, restPosition, currentPosition ) - very handy for modeling of curves and preanimating the dynamic simulation. The worklflow is very open and you are not forced to fallow it at all - just organize a new one, which will meet your needs better.
For example - attach the curves to nurbs geometry and style the hair as you want easily.

We can mimic part of this functionality in Houdini, but will need its time to be done.

just curious, can you describe their “hair-to-hair” interaction?

From experience, that isn't an easy thing to do accurately.
-k

The collisions are very accurate.
As far as i can remember the first thing i tried was to test the perHair collisions, because we already know few commercial products which totally suck in this area and i was worried that the dynamic curves are just another half baked solution.

Every dynamicCurveSolver has sampling attributes controling the collision quality, As mutch push them as precise is the simulation. It is fast. You can override the solver values on per follicle basis - local perHair control. I cant describe everything with few words. Here is not the right place to talk so mutch about Maya.

I don't think is a bad thing, talking about their fur.

There isn't a off the shelf solution in Houdini. Everyone has their own HDK code to work with hair. We have control hairs, and two level of guides, and a number of inhouse operators to handle deformation. I can't really go into detail about ours, but we do render the bulk of our hair as procedurals via Renderman. The use of procedurals to generate the hair at render is greate but at times it is very difficult to handle hair-to-hair interaction that way. Like anything it depends on the groom, and deformation of the hair. The stuff in the Incrediables was very carefully designed to hide certain things and to help certain things. It was very cleverly done, both in design and deformation.

I'm interested to see how SESI is going to tackle the topic and I think any input from people would be appreciated.
-k

Yeah, it is definitelly difficult to get proper collisions when the hair primitives are generated at rendertime.

The dynamic curves have “collision width offset” called thing, It defines the “volume” of the curve. Some tweaking will be needed to match this “volume” to the volume of the hair strand controled by the curve. The curves will start colliding when their volumes penetrate.

I agree with you, the hair is tricky thing - there is no perfect solution. The clever hair design and workflow helps a lot - the incredibles is a good example.

Interesting, I came to the same conclusion when I was setting up some stuff for our system. It can't be a fast thing to calculate.

I laughed hard when I read the siggraph notes on hair last year. By the end of paper they didn't really have a way do any of it.
It was funny.

-k

The curve solver is very fast ( relatively of course ).
The trick is to find the best relation between solver's samplingQuality and CV count of the curves.

for example: 100 curves * 50 CVs * 3 iterations = 1500 cycles ( more CVs = detailed simulation, less samples means that some curves can penetrate )
100 curves * 20 CVs * 5 iterations = 1000 cycles ( less CVs = less detailed simulation, but more iterations means no more intersecting curves )
if the CV count is too low, the solver cant do nice looking hair-to-hair collisions.

Of course, it is not that simple, but there is a big space for optimizations and tweaks.

How you are handling all this ?

I cant remember the end of the paper last siggraph, if you remind me we can laugh togather.

Hey Guys,

SideFX has Hair in the DOPS. So next release of Houdini will have a hair DOPS solver. DOPS will solve all your current limitations we have in Houdini now. After seeing the demonstration I suspect it will be the Best Solving system on the off the shelf commerical market. DOPS is truely amazing. So go ahead and make it in Houdini you will be able to do the cloth and the hair in houdini later in the DOPS. Thats what our little studio is waiting for is the DOPS. lol I know one of the problems of Maya's hair is the fact that it is a post render effect in maya software render. So it paints the hair ontop of the frame which is a real problem. I don't know if you still have this problem if you render out to Renderman. I don't know if you can render Maya's Hair out to Mental Ray or not in 6.5 Maya. I have not used 6.5 so I can not comment on it. I know in Maya 6 they still have not fully intergrated Mental Ray and Maya together and you had to enter a script command to bring up the hidden interface to the Mental Ray Shaders to shade directly in Mental Ray. Its a mess. So if you render in Maya and it renders the hair as a post effect then any transparencies, etc, etc, etc the hair just disappears. So that can be a real problem. Or you can have the hair pass through objects that it should not pass through because it was not properly occuled by various objects because it was painted ontop of the frame instead of in the actual render. Houdini DOPS would not have any of these problems. I don't know how fast it will be. I mean its totally procedural which means you have incredable flexibility but that generally means it is a bit slower too. I think the trade off is worth it though.

Cheers,
Nate Nesler

I just can't wait to see DOPS comes to the market! I also hope to see Houdini 's cloth in months!