Important Message: The CG Student Awards is now officially The Rookies.

To find out more about this exciting news please visit our new website.

Our blog has also moved to

The Making of a Demonic Warrior

Posted by on in Tutorials

This character started as a challenge. I wanted to test myself by first choosing a concept that was sketched loosely, something that I would come across quickly in production. I dug through Alex Pascenko’s amazing deviant art page. His work has enough information to work off of quickly, but also has that sketchy feel that would specifically give me the challenge of solving materials and unclear areas in 3D. 


The female demonic ogre stood out. It had a powerful pose and an aggressive and emotional look. She was also in a very open posture that would allow me to focus more time on modeling and less time on posing. Since this was for my demo-reel class at Gnomon School of Visual Effects I only had two weeks to complete it, ultimately it took me about 2 and a half weeks.





My process is to work between Maya and ZBrush simultaneously. I modeled a quick female base mesh, originally in a typical A-pose. Once the anatomy was down in ZBrush I brought it into Maya and adjusted the scale. Since ZBrush Is not a full 3D program I usually bring my models into Maya every so often for look development.




Once In Maya I set up my camera before anything is modeled so I can understand the look that I will get in the final render. Since this is more of a portrait style character, I adjusted the Focal Length of the camera to 90. This flattens out the image slightly, which is personal preference and a look that I have grown to like.





After I scale the model I bring it back into ZBrush, so my Maya scene and my ZBrush file are now 1-1 which makes jumping between software very easy. My plan of attack on this specific character was to get workable geometry quickly. So the first step was to get extremely basic geometric shapes. I would first get a generic shape in ZBrush, bring it into Maya, and then retopologize it.


The hard surface objects are generally just modeled in Maya from the beginning, then brought into ZBrush for detailing. If they are a complex shape then I do a sloppy dynamesh and get it into Maya and retopologize and work with low-resolution geometry.




Once all items are accounted for I bring them into ZBrush and start sculpting. Everything is subdivided to a high-enough level that I no longer need to dynamesh and lose my lowest subdivision level. I export all the lowest subdivision levels back into Maya so I can set up my UV’s. Each UV tile has a specific purpose. For organizational reasons, I group all my objects by material and layout my UV’s so each UV tile has a specific material associated with it. This character used 8 tiles.


I re-import all meshes back into ZBrush . I import them at their lowest subdivision so that all the higher subdivisions are re-created and projected back onto the appropriate subdivision levels. From there I exported out a normal map pass and a 32bit displacement. The only reason for the normal map pass is so I can get the illusion of detail when I paint on the low-resolution geometry in MARI. There are no normal maps in the final render, only high-resolution displacement mapping using the Catmull-Clark subdivision algorithm.




  • I group all my objects by material and layout my UV’s so each UV tile has a specific material associated with it.


In the texture painting process I import all separate materials into objects, apply the normal map to a basic Blinn shader in MARI so I can see the illusion of detail, and begin painting. Using a variety of textures, tileable procedures, and projection painting I can acquire a seamless look throughout the model. I make sure it looks good with a lambert shader on it, that way I know that if something is looking off, it has nothing to do with the diffuse texture.


Any object that is metal still gets a diffuse pass, but I only use that painted diffuse to acquire a gloss and specular map, this technique is used so I can extract any non-metal details that lay on top of the metal. I can mask them out and apply them to a layered shader once I get into tweaking the shaders. To finish the texturing process, I export all the files using MARI’S <UDIM> tag so I only have to edit one file texture if something occurs, rather than multiple file textures.





I have used Arnold for a previous project before but it was not proficient at using as a main rendering engine, so I wanted to practice on this model. My process was to first apply all the displacement maps to the model. Objects like straps and certain cloth that didn’t require a full displacement map were subdivided in Maya to speed up render times. After the displacements were all set up on a per-object level I lit the scene. I did a few passes of lighting scenarios so that I could make sure the shaders were accurate to the material they were supposed to represent and I was not being fooled by the lighting set up.





Getting accurate looking materials is the hardest process for me. Every material gets a specific shader. Since the helmet and the shoulder pads are the same material they require a single shader, so the visual consistency becomes uniform. Then the process of fine tuning all the shaders begins. I render out after every adjustment, turning displacement off, lowering resolution, all the tricks I can think of to look-dev as quickly as possible and get the results I need. Finally, the work pays off and a nice beauty render pops out.




The last step of the modeling was they hair. I used a variety of cards and the Maya Geo Hair 2 script by Thundercloud to get nice quality hair. Since the hair was so thin I made sure that my anti-aliasing quality in Arnold was 10 across the board. Once the hair looked good I applied the Arnold hair shader to overwrite the Maya shader and the hair was finally completed. I thought it was going to be the most difficult process, but It turned out to be one of the easier things I had to do and gave me no trouble.


I then set up all my render passes for a final pass in Nuke. I rendered out the hair separately and put everything gave everything a multi-matte so I could adjust everything on a per-object level in Nuke.




In Nuke I rebuilt my Beauty using all my passes, put on an AO pass, and tinted that. I tinted the AO slightly red on the skin and bluer on the rest of the objects. I then adjusted the specular, gloss, and color in real time using Nuke. If the Beauty render out the engine is pretty good, it makes adjusting everything really quick in Nuke, so I only spent about half a day fiddling and tweaking the rendered images. I renderer those out at about 4k, put it into Photoshop for a little touch up, and the images you see were completed.


I like to use this method and linear workflow so that I can render out image sequences and the same adjustment are applied to every image, rather than paint over one single image at a time. I can adjust the camera, pick a new camera angle, and all the same adjustments will apply. I also wrote Photoshop actions in case I wanted to render out a turntable and apply the same actions to multiple images. The final image just under 2 hours per frame, rendered out of Arnold at 4k, all textures were 1k resolution.




Omar Aweidah

Omar Aweidah


Javier Blanco has created an incredibly detailed breakdown of the creation process for his very first demo reel piece at Think Tank Training Centre, with Chris Nichols as my mentor. Javier tells us it...
Adrien has created a detailed breakdown of the creation process for one of the environments for a short film called Initium, Adrien was one of a team of students from ArtFX in France to complete ...
My name is Tuna Unalan and I am a CG Generalist with a concentration in lighting, rendering and look development. I am a recent graduate from Savannah College of Art and Design. I will be explaining m...
blog comments powered by Disqus