Making Of 'Le Rabbit'

Introduction

"Le Rabbit" cannot be described as the result of an extremely well organized and planned process. In most of my personal work I give priority to the "fun factor" and try to keep the whole process really enjoyable. So, I try not to plan everything in detail at the start, and tend to leave a lot of decisions to be made along the way in order to keep all the phases creative.

During the creation of this image I bumped into several problems, especially regarding the fur. I am not a digital hair expert and I had to face some questions for the first time. These problems may sound basic to some readers, but I have decided to talk about them and the way I solved them, because some of you might consider these tips valuable.

Regarding the motivation to create this piece I had wanted to make a more cartoonish character for ages, because this is a type of language with which I feel really comfortable. An image about a humanized rabbit with lots of attitude started forming in my mind ... and what was supposed to be a modeling exercise of just a few days ended up being a finished piece that took me about two months.

Modeling

From the beginning I had the intention to sculpt the model in ZBrush. At the time I was not so familiar with ZSpheres and I wanted to make sure that the character had enough detail in certain important areas, like the snout, so I decided to go old school and  create a low-poly version of the rabbit with good topology for further detailing in 3ds Max.

I created a box with a few subdivisions and applied the Symmetry Modifier on top, in order to create a mirror axis at the center of the body. I then applied the Edit Poly Modifier below the Symmetry Modifier, and started editing the polygons with "Show end result" turned on so that I could see all my editing being mirrored, to have a clear idea of the full silhouette.

The usual Edit Poly tools, like Extrude, Chamfer, Connect, etc. were used to create a T-pose version of the character. I created some extra edgeloops around the joints (elbows, neck, wrists, knees, shoulders) so that I could pose him later and have enough polygons to work with in those areas. I also decided to model the fingers individually, rather than just give volume for the hand, so that later I could position the fingers more accurately (Fig.01a - b).

Fig 02

Fig 02

Fig.01a

Fig 03

Fig 03

Fig.01b


One thing that I have learned from previous projects is that it is very important to have the eyeballs' geometry present while we model a face. It is very easy to incorrectly model the area around the eyes, especially the eye socket area and the way the eyelids make contact with the eyeballs. So, I created two spheres in the place of the eyes to guide me. With this particular character, the eyes have different sizes. I did this to emphasize the rabbit's expression and really exaggerate the lift of the eyebrow. It also contributes to the wacky feeling of the character.

The basic shape of the front teeth and the gums were also created at this stage to help me with the modeling of the snout, and to make sure they fitted well.

UV Mapping

Usually it's at this stage that I take care of the UV mapping (Fig.02). It is a good time to do it because the geometry is still quite simple and easily identifiable on the UV layout. If you leave it till later you will have to deal with thousands of polygons and the process might not be so simple.

I tried to minimize the distortion of the mapping, as well as keeping as much continuous skin as possible. The advantage of keeping the geometry mapping continuous is the fact that you can paint over large areas without worrying about the continuity of the texture between different parts of the model.

I also kept all the different parts on the UV layout at the same scale. I could have used some more texture space for the head, for example, but then I would have the problem of different pixel scale on different parts of the model. Fortunately, to deal with the continuity problems on the edges of the different parts, I could count on ZAppLink, but we will talk about that later.

For the UV unwrapping I used the Unfold3D application.

Fig. 02_model_lowpoly_unwrap

Fig. 02_model_lowpoly_unwrap

ZBrush

After setting the UVs I attached the gums, teeth and eyes to the body mesh and exported it in OBJ format, which was then effortlessly imported into ZBrush (Fig.03).

Fig. 03_model_zbrush1_import

Fig. 03_model_zbrush1_import

With the mesh inside ZBrush, I used the Tools > Polygroups > Auto Groups option to easily separate all the elements into different Polygroups. Then in the SubTools menu I have chosen Group Split to automatically place each polygroup into a separate SubTool. This way, the eyes, gums and teeth were stored as individual SubTools.

The following step - a very important one - was to define the Polygroups of the main body. By separating the mesh in different groups you can hide/mask parts of the body during the sculpting stage. This allows you to reach inaccessible areas after posing and making selections for transposing a lot quicker. With the model in a T-pose it was very simple to make the selections to define the Polygroups. I defined the polygroups according to the anatomy of the body: head, torso, arms, forearms, hands, thighs, legs and feet (Fig.04). Later, for hand positioning, I also had to separate each individual finger into a Polygroup.

Fig. 04_model_zbrush2_polygroups

Fig. 04_model_zbrush2_polygroups

I usually use the Symmetry tool while sculpting for as long as I can, in order to minimize doubling the work in certain parts of the body; however, with this project it was not the case. The pose of the rabbit was my main concern and I wanted his muscles to bulge and shape correctly according to their final position, so I took the option of posing each body part first and shaping its anatomy later.

I used the Transpose tool for all the posing, mainly by masking the part I wanted to position, inverting the mask selection, setting the Transpose Action Line as if it was an actual bone, and rotating it around the joint (Fig.05). After rotating, it is common that the joint area doesn't deform exactly as you would like, but a few brushstrokes can usually fix that. The Polygroup distribution was really handy at this time, allowing for a fast selection.

Fig. 05_model_zbrush3_transpose

Fig. 05_model_zbrush3_transpose

Regarding the sculpting technique, there was nothing new about it, I mainly used the Move tool for reshaping, the Clay tool for sculpting (I find it more controllable than the Standard brush), the Inflat tool for muscle bulging, the Flatten tool to smooth the surface (I prefer it over the Smooth tool for this purpose), and the Smooth tool for untangling the surface when deformations get ugly - particularly at the inner parts of the joints after transposing, where geometry tends to self-intersect.

I've prepared an image in which you can see the evolution of the sculpting phase (Fig.06).

Fig. 06_model_zbrush4_evolution

Fig. 06_model_zbrush4_evolution

As I have mentioned, my main concern was the pose because it defines the personality of my character. I tried to achieve a natural pose by shifting the character's weight to his left leg and tilting the hip so that his right leg could advance a bit, while keeping both knees locked, resulting in a comfortable standing position. I also tilted the shoulder line and lowered the head in the opposite direction, while the eyes keep looking to the side in order to give us "that" look. The raising eyebrow is also emphasized by the ears, as the rabbit's right ear, above the bigger eye, points up, and the left ear curls down in continuity with the spine curve (Fig.07).

The rabbit's right hand, which will later hold a cigarette, was also positioned with an exaggerated angle backwards - not only to be more expressive, but also to create a negative space between the hand and the cheek so that the silhouette is clear without any overlapping.

Fig. 07_model_zbrush5_pose

Fig. 07_model_zbrush5_pose

ZApp Link

All Pixologic's plugins are worth checking out, but my favorite is definitely ZAppLink (you can download it at www.pixologic.com/zbrush/downloadcenter/zplugins). This plugin projects the image in ZBrush's document screen into Photoshop, where you can use all your favorite tools to paint your model and then project the result back to the model texture in ZBrush!

I started by picking a white image with 4096 by 4096 pixels as my model texture. I then activated ZAppLink (under the Document tab) with Photoshop already open in the background. Choose "Drop Now" and Photoshop will pop up with the ZBrush image. The document has a layer order and naming convention that you will have to keep in order for everything to work as expected (Fig.08).

Fig. 08_model_zbrush6_ZappLink

Fig. 08_model_zbrush6_ZappLink

Add as many layers and blending modes as you want, as long as in the end you collapse all your painting to a layer with the name "Layer1" with the original mask. Then save the image and go back to ZBrush, choose Re-enter ZBrush, select Pick up now, and the painting will be applied to your texture. Don't forget to save the changes to your texture by going to the Texture tab and choosing Export.

To continue painting the model you just have to change your point of view and project the image again through ZAppLink, and keep repeating the process until you're done. You can even hide parts of the model before making the projection in order to paint inaccessible areas. This is how I painted the diffuse texture of my character (Fig.09). The remaining textures (Specular, Bump, SSS, etc.) were fully painted in Photoshop using the diffuse texture as a base.

Fig. 09_model_zbrush7_ZappLink2

Fig. 09_model_zbrush7_ZappLink2

The Fur

I exported the high-poly model in OBJ format and imported it into 3ds Max at this stage. The model had about 320,000 polygons and my new task was to cover it with fur. As you may know, when using the Hair and Fur (WSM) modifier, a hair guide spline will be created for each vertex. By manipulating these guides you will define how the hair will grow, as the software will make an interpolation between the vertex splines in order to create the hair strands in the area between the guides.

However, growing the hair on a 300K poly mesh is not an option. No one wants to deal with 300,000 guides - not to mention that the software will not allow it. The only option is to use a low-poly mesh to grow the hair. So, I exported the mesh from ZBrush at a lower subdivision level with about 5,000 polygons.

Using a lower subdivision mesh brought up another issue: the polygons of the high- and low-poly models were not coincident; some vertices were below the surface of the high resolution mesh and others were above. Having hair growing in the air at a distance from the high-poly mesh was a problem, but having the hair grow below the surface is actually desirable. As such, I applied a Push modifier with a negative value to the low-poly mesh in order to shrink it below the high-poly surface. I then turned off the Renderable option under the Object Properties, so that the low-poly mesh didn't render (the hair will render independently from the fact that the mesh will not). I applied the Hair and Fur (WSM) modifier, and the titanic fight for hair control started (Fig.10) ...

Fig. 10_hair_placement

Fig. 10_hair_placement

With this project I found that the best way to control hair is to shorten it drastically and get rid of all the randomizing parameters at the beginning. Here are the steps of how to do this:

  1. Go to Frizz Parameters and reduce the Frizz Root and Frizz Tip values to zero
  2. Choose Style Hair and increase the brush size so that it covers the whole model
  3. Remove the Distance Fade option; pick the Scale option and drag-and-click to the left over your model, until the hair guides are really short
  4. In the Utilities box, click on the Recomb button so that the hair falls along the surface
  5. After, you can choose the Stand option inside the Styling box, and drag-and-click to the right slightly over your model to raise the hair guides a little
  6. This should give you a good starting point. After that, grow and comb the guides in small areas at a time, hiding the rest of the guides
 
It is a work of patience ... lots of it!

Another thing I learned during this project was the obvious fact that fur adds volume to the character (Duh!). All of a sudden, the nice sculpting details were covered by a thick layer of hair. I had to invest a lot of time combing the guides to make the fur flow along the skin's surface and keep a sharp silhouette. Fig.11 shows the evolution of the combing, as well as the lighting.

Fig. 11_hair_test1

Fig. 11_hair_test1

It also came as a problem that the hair didn't look the same when rendered at different resolutions; it was very difficult to spot mistakes and make decisions on the correct density at low resolution renders. In order to be able to test the hair correctly, I created a test scene with a single light, without any shadows or GI, and just a simple diffuse material for the skin mesh. Most of my hair render tests were made in this scene at about 3,000 pixels, and with render times of around 2 minutes.

After having reached the conclusion that I needed a hair count of about 2,700,000 hairs (!), another problem came up: I just couldn't render anything larger than the 1,500 by 3,000 pixels in resolution. I used the "hair buffer" method for hair rendering ("mrprim" and "geometry" couldn't handle this amount of hair), and at the time of rendering only two-thirds of the hair showed up. Later, I discovered that there is a 70MB buffer to write the hair to, and at higher resolutions this buffer value is not sufficient. In order to fix this, go to Effects > Hair and Fur; inside the Buffer Rendering Options increase the Tile Memory Usage, and that should solve the problem. The bigger the resolution, the bigger the buffer! It sounds simple when you know what to do...

In Fig.12 you can see the three different base meshes I used to grow hair on the body, snout and tail, as well as the hair count for each. The hair guides are represented in yellow.



Fig. 12_hair_lowpolymesh

Fig. 12_hair_lowpolymesh

In Fig.13 I have compiled all the numbers used to generate the final hair. For the Tip Color and Root Color I used textures with a brownish tone - brighter for the tip and slightly darker for the root. Don't forget that the color swatches for the Tip and Root multiply by the texture color, so change the swatches to white or else it will affect the texture color.

Fig. 13_hair_settings

Fig. 13_hair_settings

Fig.14 shows the hair resulting from the guides and settings previously described.

Fig. 14_hair_test7

Fig. 14_hair_test7

The Skin

The renderer I chose for this project was V-Ray, a long time companion.

The 1.5 SP3 version of V-Ray presents a new excellent shader, the Vray Fast SSS2. It was created to make subsurface scattering materials fast and easy. I have used this shader for the first time in this project and it is very similar to Mental Ray's SSS Fast Skin Material. I can confirm that it is blazing fast!

Subsurface scattering was crucial in this piece because I had set up a strong light from the back and it allowed me not only to emphasize the rabbit's silhouette but also to represent the characteristic translucent effect in the ears. You can check the used textures and shader settings in Fig.15, as well as the render result of the applied skin material in image Fig.16. You will notice, by examining the fingers on the right hand, that the subsurface scatter radius is a little bit exaggerated. However, the important parts were the ears, as the rest would be covered by hair.

Fig. 15_skin_shader

Fig. 15_skin_shader

Fig. 16_skin_test2

Fig. 16_skin_test2

The Eyes

For me, the most important part of a character is the eyes. Unless you are representing a dead corpse, if the eyes don't look alive, you will ruin the entire piece. People look at each other's faces all the time and the main focusing point are the eyes. We are so used to observing eyes that if something is wrong in their representation, we will immediately spot it.

I believe that there are the nine essential points in eye representations that have to be nailed in order for your character to look lively (Fig.17), as follows:

  • A. Shadow line below the eye lid falling on the iris
  • B. Bright highlight revealing the strong light sources - I used an HDR image on the environment to achieve this
  • C. Highly reflective cornea (not forgetting the Fresnel effect)
  • D. Reflection/brighter area on the iris on the opposite side of the main highlight/reflection
  • E. Smooth dark edge surrounding the iris
  • F. Recessed Iris - the iris is not so recessed in reality, it is actually quite plane. However, by pulling back the pupil we emphasize the refractive effect of the cornea and the liquid that exists inside the eye, instantly looking more alive
  • G. Solid cornea - it is very important to model the cornea with thickness so that you can use an Index of Refraction and slightly distort what is behind it. In this case I have used an IOR of 1.4
  • H. Milky sclera - the white of the eye has a milky look and shouldn't be represented as bright white. I used the VrayFastSSS2 shader with subsurface scattering to achieve that look
  • I. Wet look - the tear line in the contact area between the cornea and eye lid. I created a spline with thickness, converted it to Editable Poly, and changed its shape in order to achieve a watery look

Fig. 17_eyes

Fig. 17_eyes

Composition

One trick I use to test the composition of a scene is "paintovers". I take an image into Photoshop and extend/crop the canvas size, twist it, paint new elements, change colors, etc. It is a quick and dirty method to preview changes. If I like them, I recreate them in 3D.

Fig.18 is one of those tests. In this one I took some decisions, like placing an ashtray near the character and a few framed paintings on the back. I decided to flip the image horizontally because it worked better that way - we read it from left to right. I also added a slight tilt to make the composition less static. The rabbit's eyes are at a one-third distance from the top and from the right side, in order to become the center of our attention.

In the final composition (Fig.19) I had some trouble making sure the Mona Lisa did not compete with the rabbit in terms of attention. For that I had to let the rabbit's ear slightly overlap Mona Lisa's frame, so that it became obvious that the painting was behind the rabbit. I also intentionally left Mona Lisa's eyes out of the picture to avoid having two pairs of eyes fighting for the viewer's attention.

Fig. 18_composition_test

Fig. 18_composition_test

Fig. 19_composition

Fig. 19_composition

Compositing

Photoshop was used for all the compositing. I had already done a lot of lighting and material tests in 3ds Max, and as such I didn't find it necessary to create render passes to fine-tune the lighting, reflections, etc. further.

I rendered three separate hair passes - body, snout, and tail - by setting the Composite Method to None in the Effects > Hair and Fur window. I rendered the base image with the hair shadows but without the hair by setting the Composite Method to Off in the Effects > Hair and Fur window. A ZDepth pass was also rendered and used to defocus the background with the Lens Blur filter in Photoshop. In addition to this I rendered an Occlusion pass using Mental Ray, which I consider to have a far superior occlusion shader comparing to VRay's dirt shader (Fig.20).

To finalize the image, I painted some dirt marks on the wall and on the floor, painted the smoke in the air and coming out of the cigarette, and added more light from the top right corner by painting over the image with the Color Dodge blending mode.

Fig. 20_compositing

Fig. 20_compositing

Conclusion

The technical aspects of a piece are important, but what really matters is the reaction it provokes on the viewer. I like humorous images that make people smile. That was the main goal of this piece. I also introduced some secondary comic elements on the image, like the smoking Mona Lisa and the carrot earring, because it's fun for the viewer to discover the jokes as they explore the image.

I am very happy with the way the final image turned out (Fig.21). I hope you have enjoyed this article as much as I have enjoyed writing it for you. Thanks for reading!

Fig.21

To see more by Jose Alves da Silva, check out Digital Art Masters: Volume 9 and ZBrush Character Sculpting