Making of 'Female Rogue'
Concept & References
In a project like this one, I always like to find as many references as I can and to be very clear about what I want to do before starting to model. For this image, as the subject was a rogue, I searched for references in videogames, illustrations and real photos ... I always try to get a large amount of them, and then I try to design nice clothing, a good pose, and so on. For example, the body armour was inspired by a lingerie photo (Fig.01); I used it as a reference and started tweaking it to make it look like an armour. I also got the idea of that a pattern from there, too, which I used on some other parts, such as the small shield and on the skirt.
I used 3ds Max and ZBrush to model. I did everything in a T-pose at first, instead of modelling it in the final pose, as I thought it would be nice to have the possibility of animating the body (even though I'm not planning to do it) (Fig.02a - b). I started creating a low poly version of the body. In this case, I did the body and the head on different meshes, because I reused a headless body which I had already done for an older project, but ideally it would have been better to model them together to avoid the merging of the body and head parts.
Â Â Â Â Fig.02a Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Fig.02b Â Â Â Â Â Â Â Â Â Â Â Â
Anyway, the process was pretty much the same; once I had the low poly version, I took it into ZBrush. I always work the same way here: I subdivide the model once, and then I do as many details as I can in that subdivision level. Once I'm happy with that level I subdivide it again and continue detailing until I get around 7 subdivisions. I try to do as many adjustments as I can in the lower subdivision levels, because the more detail you can do there, the less you will have to tweak in the higher levels; it's trickier to keep the model nicely smoothed if you have to do big tweaks in the higher levels. Once happy with the results I went back to Max, using displacement and normal maps to get the details from ZBrush.
Half of the clothes were done using the same method (modelling the low poly version in Max and detailing in ZBrush) and the other half (the more rigid ones) were modelled in Max only, as I thought it would be faster that way. There's nothing special about them; I usually duplicate the body mesh, delete all the parts I don't need for the cloth, and start modelling from there using the body as reference to see how the clothing should look.
When texturing, I always use a combination of real photos and handmade paintings to create them (Fig.03a). I usually create a base texture using real photos and start adding more layers over them, using more real photos and/or handmade paintings to get different colourations all over the texture. I use that method in all kind of textures - body, clothes or environment. Using the body texture as an example, I start creating that base layer using real photos to cover the whole texture map, trying to delete all the details that are not needed, such as the highlights, shadows, wrinkles ... Then I start adding layers over it, using real photos to add details (like the lips, nipples, belly button and so on), using hand painted layers to add different colorations (adding reds and blues, making darker and lighter zones and so on) and also adding smaller details, like wrinkles, spots and veins, etc.
With all those parts on different layers allowed me to create the other maps I needed (specular, bump and so on) very quickly, adding a few hue/saturation or brightness/contrast effects on the top layers, and I was also able to turn the layers on and off when I didn't need them for each texture.
I used Mental Ray to render the scene, so most of the materials were done with the Mental Ray architectural material. There isn't much to explain about this; most of them had diffuse, bump, specular and environment maps, but perhaps the skin material is more interesting (the material can be seen in Fig.03b - c). From top to bottom, the maps are the diffuse map, subdermal map (a much redder version of the diffuse), the specular map and two blue noise filters to give some randomness to the specular parts. There's also a normal map in the bump channel, but this can't be seen in the image.
Posing, Environment & Lighting
I did the posing in Max, using a biped to move a the low poly version of the body and using the skin wrap modifier in the high poly objects so the biped was moving the low poly version and the low poly version was moving the high poly objects (Fig.04). Once I had the pose I wanted, I just had to tweak the parts that were not looking too good, like the shoulders, elbows and so on. I did this using a proxy version of the environment, a few boxes and planes - just enough to know where the volumes would be - which also helped me in placing the camera.
Once I had the body posed, I did the high poly versions of the environment. Most of it was also planes and boxes (the walls and the wood parts). The only part which was a bit trickier was the roof. To do this part I modelled and unwrapped one of the tiles and then I cloned it to create a vertical line, making sure to do the same with the unwrapping, keeping all the pieces in a vertical line and not all in the same place. Then I duplicated all the tiles, mirrored them and placed them to create one of the lower lines. Once I had that I duplicated the entire structure again - enough times to complete the roof. This way I just had to tweak the unwrap again, placing the unwrapped tiles more or less in the same way they were placed in the 3D scene. With the previous work I had done I just had to select each of the lines and move them, instead of having to move them one by one. Once all this was done I used an FFD box modifier to get some randomness on them (lowering some parts and moving a few others up), and I finally moved a few of the tiles individually to finish adding a bit more randomness.
I only used Mental Ray Spot lights and Omnis to light the scene. There were a few small bluish sources of light in the outside environment, a bigger whiter source to better light the character, and strong yellow lights from inside the building. I "cheated" a bit in some parts to achieve better lighting. For example, I used two yellow Omnis only affecting the character to add more yellow light coming from the back of her, to get a nicer profile and some contrast with the background. With only the light coming from inside the building that effect could not be done, so I added those 2 lights placed outside the window to win that effect - even if it wasn't physically correct.
Hair is always a tricky part to do, and it has been is a special part in this project, too, as I did lots of tests with different methods until I found one that worked (Fig.05). First I tried with the hair plug-in that comes with Max, but I was really unhappy with the results (the same happened every time I tried it, but for some reason I kept trying again and again). It was really slow to render, I was getting lots of random errors that were resetting all the combing I had done, and the result wasn't half as good as I expected...
The second method I tried was the old textured planes method, and I got a decent looking hair in the T-pose model, but once I rendered it in the final pose with the final lighting it didn't look too good...
I had almost finished the hair with the textured planes method, but I decided to give the Hairfx plug-in try to (or Hairtrix, as that's the last version where you have the Hairfx and Ornatrix plug-ins combined). The results couldn't have made me happier; I was able to create the new hair starting from zero very quick and easily, and the result was very nice (or at least that's what I think!). Considering the small amount of time I spent doing tests with this plug-in (as I wanted to finish the image as fast as possible at that point), I can't wait to use it in a new project to see what else can it do!
There are some Hairfx tutorials that explain how to create hair with it better than what I can explain here, but I can share a couple of tips I've learned while doing it:
First of all, even if the last versions of the plug-in can render in Mental Ray, the render times become crazy, so I found it much easier to render the hair in Scanline in a separate pass, and then combine them in Photoshop.
Finally, with all these kinds of hair plug-ins, I always try to create the hair using different hair settings and growing objects for each zone. This gives me much more control on the hairstyle than if I try to create all of the hair growing from a single mesh. Apart from that, the process for creating hair is easy - you just have to keep creating growing zones and hair guides to define the hairstyle you want!