Samurai Iko
Xabier Sevillano talked about the production of his character Iko during CGMA course Character Creation for Games mentored by Patrick Yeung.
Introduction
Hi, my name is Xabier Sevillano, I am a 3D Artist from Spain. As far as I can remember, video games have always been part of my life. I remember a teacher asked me once, “What do you want to do in the future?”. Since in my town, there is not much choice and most people study some kind of engineering, I answered, ”I don’t know, engineering”. He replied, “Why don’t you study for making video games?” because he saw me playing on my PSP every day. Funnily enough, until that moment I never thought about it. Could one make a living making video games?
I studied computer programming for 3 years, but in 2011 Digipen Institute of Technology Bilbao opened and I enrolled aiming for the 3D Art bachelor. This decision was made because I got acquainted with ZBrush 4 while studying programming and I fell in love with it. I wanted to do 3D Character Art like all those guys from ZBrushCentral. My parents didn’t understand the decision, Digipen’s counselor didn’t understand it either, but I was happy.
In 2015, I finished university and joined Davalor Salud, a company that was doing VR video games for checking the health, as a 3D Generalist. In 2016, I decided to change my life a little, so I took a risk and went to Vietnam to work on VRnam, a project for teaching commercial airplane pilots in VR. There, I worked as a 3D Artist. After working for 3 years there doing 3D cockpits (hard-surface modeling 95% of the time) I decided it was time to chase my real goal to be a 3D Character Artist. So I decided to leave the company and start creating a 3D character portfolio. I wanted to do it the best I could, that’s why I joined CGMA’s Character Creation for Games course with Patrick Yeung.
Iko
Iko: Character Origin
At CGMA, they show you how to create a video game character from 0. There is not enough time to do a full-body character so they only require a bust from you. I knew I had the time and I wanted to make the artwork as good as possible since it was a piece for my portfolio, so I went for a full-body character. After looking at tons of designs on Artstation I decided to go for the one from Hou China. It was perfect for my purpose, simple but with a lot of big shapes. Also, who doesn’t like cool samurai demon warriors with huge swords?
Concept art by Hou China
The name Iko comes from Nariko from Heavenly Sword, one of my favorite games on PS3. Every time I see a red-haired girl it reminds me of Nariko – the same happened this time when I saw Hou’s concept. Also, I was not sure what the name of his character was (his series was called Ji He Zi but I was not sure if it was the name of the girl or the story). So for that moment, I decided to call her Iko and find a name later on. As you can see, the name stuck.
I am text block. Click edit button to change this text. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.
Modeling Body
As always, the first blocking is really important so I gave it a push and really put effort into making it work from the beginning. The approach to the blocking was basically using subdivided cubes to make everything. I used this method before but this time, the number of objects to model was way bigger. Anyway, I really enjoyed this part, since you start giving shape to your character and any addition looks like a huge leap forward.
Once I established the blocking, it was really easy to make changes and move things around. After that, I used Dynamesh, started adding details, and kept developing the character. This is one of the slowest parts of the process and with so many details in this character, it was a little bit overwhelming. One of the things that helped me a lot was making corrections and notes to my future self. Before going to sleep, I took 10 minutes to analyze what I did during the day and made some corrections. When I woke up the next day I had already a guideline to follow. Sometimes, I disagreed with my past self, but it was a nice way of keeping track of my thought process.
In the beginning, I was planning on going a little bit more realistic, with a lot of cloth details, a realistic face and hair with hair-cards. After checking how much work I needed for the hair-cards, I decided to drop that idea, I had enough work already with the model. So I started looking for other looks that could match the hair like the one in the concept. I was somewhere between Overwatch and Apex styles (referring to the sculpts on Artstation by such artists as Sam Sun, Gary Huang, or Patrick Yeung, not the final look in the video game). I remember studying Gibraltar’s stylization shared by Gary Huang helped me a lot at that time.
Overall, it was a really nice piece of advice from Patrick to change the face since, in the concept, the girl looks like quite a generic video game character.
Having the references always on my second screen or in PureRef was vital when trying to keep to stylization. My head tries to do natural modeling rather than stylizing the volumes, so I always compared my model to the references.
Clothes & Swords
Since I had some time for research, I wanted to try out Marvelous Designer. It turned out quite simple to use, although I didn’t do anything complex. I only wanted to make a base there, the rest was sculpted on top of it inside ZBrush. After importing the base into ZBrush, I did a Projection and a clean up of the pants. Later on in the process, I saw that the pants still looked too realistic, so I did retopology, reprojection, and changed the proportions a little bit.
For the armor piece on the shoulder, I didn’t have any reference since in Hou’s concepts, it was sketched only roughly. I looked at some other artworks he has done and tried to design the piece myself. I kept doing some variations until I found the final one. It was a hard task to decide which one looked better on the model and complimented the overall shape, – I really liked that aggressive shape in the concept, but my first tries were too realistic and didn’t match the leg armor style. In the end, I looked back at the concept and went for a less usable, yet closer to the concept look. Still, I tried to make the armor believable.
Different shoulder armor tests, the last one is the final:
The swords were left behind until I almost finished the project. I was not sure if I was going to have time to finish them, so I decided to leave them for later on. Luckily, in the last two weeks, I had the opportunity to finish those props.
Baking & Texturing
Before this project, I used to bake in Substance Painter. I would create a file for baking the Normals with an exploded view of the objects so that they don’t intersect and another one for the AO. However, when you have so many objects it is really expensive to maintain. If you need to change an object you have to do it in 2 files and export it.
Following Patrick’s guideline, I tried baking in Marmoset Toolbag exporting each object into a .fbx and creating different Baking Groups. In the beginning, it might be slow, but it is way easier to maintain. Marmoset also lets you modify the cage in real-time, so I’ll definitely use this software in my baking workflow from now on.
For texturing, I used Substance Painter. I used PBR for the whole model except the hair that has a SpecGloss shader to get red highlights. I use object ID to add a Base material where I need to, modify the color, metal, roughness, and height, and then add another material for some highlights in the corners usually. Then, I add more details, such as wear and tear, color variation, rust, dust or dirt.
The most interesting texture to work on in this project was the skin. I have never made “realistic” skin before, and it was really interesting to see how much different layers of micro details, skin pores, scattering, color variations, etc. can change the look.
Working on the hair was also quite interesting. I had to add a lot of color variation to the chunks of hair for which I used HSL Perceptive filter with the ObjectID map. Then, I added more variation with the thickness map and some gradient filters, and finally, I manually painted some hairs to add another dimension. Still, the hair looked like a helmet or a part that didn’t belong to the character. In order to merge the hair and the head together, I painted a shadow line on the head and darker strands on the hair.
Last but not least, Anisotropic which is probably the most important map for the hair. I used Krita, a program I’ve never used before, in order to create a direction map. It is incredibly easy to use and I’ll now use it in my workflow, too.
It is also important to go back and forth between Substance and Marmoset (or any other rendering tool) to test the textures since there are going to be a few differences. This was more obvious with the objects that had subsurface scattering or transparency, such as the skin and hair.
Presentation & Improvements
I tried some crazy lighting situations with different moods, but in the end, I always went back to common 3-point lighting. I wanted the character to speak for itself.
The most important part of the presentation was the feedback. Apart from Patrick’s feedback, I talked to some friends with different specializations (concept art, 3d art, and animation) and their feedback changed Iko completely. I think the pose was the biggest improvement. Some other changes I did were the saturation of the hair and eyes, the makeup, and the shape of the face.
Comparison sheet with slightly different hair tones:
After texturing, the shape of the face flattened a little bit and didn’t look as sharp as in ZBrush. So I went back and changed the geometry. I thought it was going to be tricky to change it at this stage, but it was easier than I expected. I exported from ZBrush, and Substance handled it perfectly – I didn’t have to do any fixing, just baked the textures again. Funnily enough, this is a small detail that nobody would have probably noticed.
Pose comparison; on the left – before feedback, on the right – after feedback:
Feedback
Studying at CGMA helped me a lot to improve. I already followed a similar workflow before, but thanks to this course, I had someone giving me feedback and advice on how to work better.
The biggest lessons learned are:
- Document your process and reflect on your work.
- The blocking stage is sacred. If you put effort here, your future self will be happy about it.
- Bake in Marmoset.
- Even at the later stages of the process, don’t be scared of changing core things.
- Feedback, feedback, and more feedback. Get it whenever possible!
