Making Of 'Sad Alien'
This is my little sad alien and I want to share with you the story of how he was born. It was in 2008, when I was taking online classes with the famous Alex Oliver. One evening I didn't have any homework from him, but I wanted to sculpt anyway. I found a great concept by Carlos Huante (Fig.01 - Character Design © Sony Pictures) and with that my journey began.
I started modeling in ZBrush (version 3 at that time, I think), from a simple sphere. It was a very rough and quick model, but in that moment I decided that I want to create a scene with him in it.
You can see the rough model in Fig.02. He seems a little bit angry at this point, but he developed into a very sad character during the creation process.
I used a standard approach, with the Clay and Standard brushes. This model took me about 30 minutes, and then I put it to one side for a few months. When I came back I started working on his body, starting with a very basic base mesh (Fig.03).
From that point on, I continued with this workflow of using simple base meshes or DynaMesh to create rough shapes and forms as it allowed me to quickly get my ideas onto the screen. When you're about halfway to a final model, and you have the proportions and major details locked in, you should retopologize the model even more and if needed, make it ready for rigging. You can see my final model in Fig.04 and how the cloth has been extracted from the body mesh.
Retopologizing and Final Mesh
There are a lot of ways to retopologize. While I was working on the alien, I was searching for the best tool. I used ZBrush for retopologizing the head, but I wasn't satisfied with its workflow. So for the body I tried TopoGun, which has a great user interface and was very quick. It was the best topology tool for me until modo 601 was released. Nowadays I'm using modo for both base meshes and retopology.
Here is the final mesh I used for the render, compared to the original meshes (Fig.05), and the extracted cloth and its retopology (Fig.06).
After the alien model was finished, I started to think about the scene. In the beginning I just wanted to render him untextured with a simple background. Then my friends convinced me to make a serious image with him in it, so I started to think about a more complex scene.
I wanted to show the audience why he is so sad and so I can up with the idea that he was an action figure, missing some part - an arm, in the final image. I wanted him to look semi-realistic and for the viewer to not be sure if he's alive or not.
I decided to place him on my desk, so I started modeling a bunch of equipment (Fig.07). At that time I was moving from Softimage to modo, so some stuff has been modeled in modo and some in Softimage. I have to say that modo is a paradise for modelers. It has much more complex tools and the modeling workflow is much faster than other applications.
You can see the whole scene in Fig.08.
In this phase I was also looking for the right composition and scene orientation. I tried about five camera angles and different object placement combinations. The time between the different versions of the scene was sometimes very long. As you know when you come back to a WIP after a few weeks or months, everything seems wrong and so you start reworking it. This is what happened in my case. I spent a lot of time changing the scene and also the lighting, but I'll talk more about that later in the lighting chapter.
Texturing was pretty easy on this piece. I made some overpaint tests in Photoshop, trying to decide how the alien should look. I decided to make his head realistic, like human skin, and his brain a little bit awkward and translucent.
I spent a lot of time thinking about what his uniform should look like. I tried different colors and materials. The green you can see in the final image is something I settled on about a week before finishing him; until then he was in a simple white dress.
For texturing the skin, I used a great hand-painting approach taken from Scott Spencer's book about ZBrush. To start I polypainted skin temperatures (red areas around the mouth, blue around the eyes, and yellows where the bones are close to the skin). Then I painted a "mottling" pass, which is something like the fiber structure under the skin. Then came the veins and finally I blended everything using a neutral skin color. You can see the results in Fig.09.
The alien's skin surface and all the details were generated from the final ZBrush sculpt (Fig.10). I also rendered out a Cavity map to add detail to the textures.
The rest of the scene is pretty simple in terms of texturing. The toy box was unwrapped and completed in Photoshop; the rest was done using photo textures and basic planar mapping for adding detail or cubic mapping for procedurals. The composition of some of the textures can be seen in Fig.11.
Composition and Lighting
The lighting and scene composition were the most difficult parts of this work. I couldn't decide what I wanted from the scene. I tried to make a daylight setup with monitors on, but it was quite boring (Fig.12). So then I tried a different composition with night lighting, in portrait orientation, but I realized that it wasn't a suitable format for the scene or as an example for a future showreel (Fig.13).
I used grayshade materials for preparing my basic lighting; by doing this you can better see light colors and intensities, and you're not disturbed by material colors. For the start I used a key light for a lamp and a rim light for another lamp. For the key light I used area and I modeled a basic lamp shade, to have exact shadow and light placement, which I wanted. For the rim light I used a pretty intense spot directing right onto the alien's brain, to support its subsurface scattering and shape.
After placing those lights I was trying to think like a director of photography. The speaker was too dark and so I added a spot light pointing to its upper part. Also the cables were not visible and the left part of the scene was a little bit dark, so I added another spot. I also added some fill lights to add to the ambience. The final scene contains about six lights (Fig.14) and a reflection box to add nice spark to the alien's eyes. I also used global illumination to add a little more realism to the scene.
Shading and Rendering
The shading was pretty easy, except for the alien. I was learning modo at the time, and I was a little bit confused by its approach to shaders. But after a while, I found it very useful and effective. I was working in modo 501 and skin shader wasn't implemented until 601, so I made my own skin based on mental ray's one. Simply adding four layers of textures with different colors and sss preferences, I generated those textures from my polypaint. I also made a slightly different setup for the brain material and seamlessly blended it into the skin using a mask, which is a very easy and effective thing to do in modo. You can see the render tree for the skin and textures in Fig.15.
The rest of the shaders are really simple and I used modo's presets heavily. I found the shader I needed, applied it on my model and made some changes to make it suitable for my scene. I also added my own shaders to the library, so I could quickly reuse them.
For almost all the materials I used Normal maps as I've found that surface bump is very important when it comes to making your scene real and using Crazy Bump for generating Normals maps from photos is incredibly easy and very effective. All the materials can be seen in Fig.16.
As for the render setup, I said before, I used modo. It was quite easy and nothing special; the only thing I will mention is that I used a 25mm lens and as I wanted to make a pretty depth of field. For this type of scene is almost impossible to use post-processed zblur based on depth, because you'll get odd artifacts at the edges. There are two solutions. Once, separate your scene into a couple of layers, and render and blur them separately, or two, simply use a brute force render DOF.
I chose the second option, so I set antialiasing up very high to get a clean effect. I also used lens distortion directly in modo, as it helped the final image a lot because a 25mm lens distortion is obvious in the real world. You can see my entire render setup in Fig.17.
The final render was in 6000 x 3374 with the render time around four hours on a hardcore workstation setup. But it's great if you consider all the render effects, SSS, GI, a lot of lights, etc. I think that modo did an awesome job and I am quite sure that other renderers would be slower.
For the last five years I've been working in the VFS industry, so post-production is my daily bread. Like a lot of the world, I use Nuke for post-processing and for this scene it was fairly simple, as you can see on Nuke's tree (Fig.18).
I always render to 32bit EXR format to have full control over the render in Nuke. My passes were just beauty and ID mask. I made a few color corrections and added local masks to highlight some areas. Then I adjusted the glows on the LED diodes and smoothed a little glow over the whole image. I decided that it needed a little bit more atmosphere, so I decided to add some dust particles. The simplest way to do that was importing the camera to Nuke and using its particle system with proper ZBlur to create the dust. I also mixed it with a dust photograph to make it more real.
Finally it came time to add a well-known combination: chromatic aberration, vignette and grain. I always caution people to use this combination wisely as it can help your image, but can just as easily destroy it. Just observe normal photos and that should help.
Here's a comparison of the pure render and the final image (Fig.19).
And here's a larger version of the final image (Fig.20).
Finally here comes the end. I have special relationship with this poor alien; we spent about five years together from the beginning of this project to the end. I'm really happy to see him finished - there's one less skeleton in my wardrobe. I hope you've enjoyed my Making Of and feel free to contact me with any questions at firstname.lastname@example.org.