Secrets of Swimsuit Babeby Jimmy Chow, Hong Kong
Rigging
The purpose of rigging is to let me pose the character rather than animate it. So the consideration is biased for realism rather than mechanism. For example, rotation plane is not considered at all. I tried to imitate the actual bone structure and so there’re joints for scapula, two for the knee, shoulder and hip. Some of them are actually not necessary, technically speaking, because they don’t rotate at all.
I just followed the general rules for joint placement. For example, don’t place the backbone too close to the back, knee joints are close to the front, elbow joints are close to the back, head joint is under the ear, neck joint is close to the middle, etc. If you’re unsure, do some tests to see if the deformation looks good. Below image shows the skeleton and the pose.
It doesn’t matter how you bind skin as long as it’s smooth bind. There’s no need to perform detailed painting for the skin weight. Use the smooth button extensively to smooth out the weights.
I used keyframes to save poses. Go to bind pose, set time=1, key rotations of all the joints, set time to something else, pose the character and then key joint rotations, repeat this process to save more poses.
No matter how you bind the skin, the shape has to be tweaked after it’s posed. The technique of negative blendshape was employed: duplicate the posed shape twice, add them to the blendshape node, set their weight contributions to become 1 and -1 respectively, tweak the one those weight is 1. In this way, you can tweak a posed character. Fig. 2.7.2 shows the pair of targets which fixes the right elbow. You can see in the time slider that time=1. The character on the left is at its bind pose. The two shapes on the right are duplicates of the target pose. The one at the back is the negative one whose elbow is unfixed. The fix applied to the shape in front is reflected in the character on the left.
Hair
Solution
It must be emphasized that the objective here is to make long hairs, not short hairs like fur. Maya Fur and Shave and Haircut are already good short hair solutions. But long hair is different. That’s why there is Maya Hair since version 6. But I’ve found it very bad at rendering as shown in the images below:
The left one was rendered by Maya software. There’s an obvious anti-alias problem on the left where the model is in front of the hair. The right one was rendered by mental ray. The hair appears hazy and it’s unable to achieve the result obtained by Maya render even if you can tolerate long render time by raising sample level and decreasing contrast threshold.
Without good software, the only way is to paint textures which most people do. But I wanted to avoid it so I came up with the idea of combining Maya Hair and Shave and Haircut, taking advantage of their own specific strengths: use Maya Hair to create hair curves, then use Shave to render the curves. The result of the experiment I did shows that this solution is good enough for my purpose as you can see in the final image.
Hair Modeling
Hair is modeled after the character has been posed. The approach is to model hair clumps using NURBS patches. In order to facilitate hair generation process, the patches are separated into four layers, namely, front, side, back and outer layers. They are shown in the first row in below image.
The front layer contains more patches so that more clumps can be seen in front and they appear thinner. The side layer includes those which are running from above the ear down to the shoulder. They are wider for the sake of reducing number of patches. The back layer is a single patch as it can hardly be seen from the camera. It is just to cover the remaining area. Unlike the other layers, the outer layer is not to model clumps. It is to model those hairs which stick out from the trunk of the clumps. These hairs are thinner and rarer. So the patches in this layer are all over the place as shown in above image.
|
|
|
