The main character took a long time of 5 months to complete as I balanced her with other production roles. A temporary model rigged with a HumanIK skeleton was created early and used throughout the project. I was able to build a base mesh within the first month of August, apply animations and test materials in September. There was a long break where I didn't touch the character art and in January it did not take long to finish the final model as everything was set up.
There's always revisions in production, so this workflow made sense. I'll cover some of the challenges and compromises in this post.
 |
| Finished character model, Marmoset Toolbag 2 |
Challenges:
Creating a convincing outfit was difficult both in design and technical implementation. As far as I'm aware, there were no female detectives in the 1920s and 1930s. There were only a handful of female police at the time and they were not given the typical patrol tasks. Despite the low statistics, Alice Stebbins Wells became first LAPD cop in as early as 1910 and helped to organize the International Policewomen's Association in 1915. With this project set in an alternate world inspired by history, it was nice when there were facts I could base the designs on.
One fact I couldn't work on was dresses which pretty much every woman wore in the early 20th century, unless they were Katharine Hepburn. Early on I struggled with convincing dynamic cloth using Nvidia's APEX tools. It was too buggy and I did not wish to get too involved with character technical art. I settled for temporary 'leggings' which were used on my temp model until I got around to finishing the high-poly sculpt. In the end the pants remove a bit of the authenticity of the design, but as a design decision for a game with lots of locomotion, it works out in my opinion.
 | |
| Left is the temp model used for half of production (just base meshes with solid materials) Right is an APEX cloth test |
 |
| Early on I wanted to play with more drapery to fit the atmosphere, but in the end static clothing was easier to handle |
 |
| Dresses just didn't work. Cut |
Influences:
In the early stages of the project I considered incorporating steampunk into the design, however that idea was cut as adding too many subgenres would confuse the main stylistic influence of film noir. I was already stretching it by adding crime and monster genre, but they were easy to fit into the narrative while steampunk was too 'out there'.
The following references influenced the costume the most. Since the lead character is a police detective, I went for more of a street cop/police look
 |
| The image on the left inspired the costume the most, especially with gun placement. |
Characterization: Designing the Face and Hair
While the clothes make the man (or woman in this case), their face and hairstyle was equally important for getting across personality. The head sculpt was actually the first asset produced for the entire project. In fact I had the look down to the materials finished before the first month of production. It seems the common methods to produce realistic characters these days is 3D scans, but I did not have access to any at the time. I used http://3d.sk/ for texture reference.
 |
| Early lighting tests in Marmoset Toolbag 2 let me study materials before even starting the game |
Since the project started in the 1920s and was changed to 1930s later, I was looking into a Louise Brooks type of bob. In the end I went for a more glamorous Hollywood style cut and the blonde coloring gave her more innocence than before. While the game is black and white, color comes through in contrast and there was just too much darkness in the game for black hair to stand out. Blonde hair is also more innocent and reflects her rookie status.
 |
| Left: Louise Brooks's 20s style didn't work. Explored more iconic Hollywood styles in the Center and came to final at Right |
Production: Texturing:
Sculpting characters is pretty straight forward. Basically have a good base mesh and make it look as legit as possible in the time given. The better the sculpt is, the best the bakes will turn out and texturing will be easier. I didn't have much detail in my sculpt due to time, but I had a clean workflow which made texturing easy, especially with Quixel Suite.
Quixel's DDO has a nice feature called DDO look dev which is a fancy way of pre-planning materials on a flat bake of your model before retopology. Color ID's are a big deal with modern texturing workflows, so I had a nice color coded model in Zbrush. I was able to decide my final look before touching a single poly, how cool! It's a bit easier to show with pictures:
 |
| 1) Separate color ID's used on each material. Polypaint in Zbrush |
 |
| 2) The color ID was baked on a flat plane as well as normal, AO, standard stuff. This is what a bake would look like |
 |
| 3) The results can be tweaked in DDO and saved as a preset to be added in an instant to the future low-poly model. BAM. |
Conclusion
There's still retopology, rigging, blendshapes, setting up animgraphs in UE4, retargeting animation and all sorts of time consuming tasks I don't have time to cover here. As I found out, while creating cool character designs and models is time consuming, it's only a small fraction of the work. That's why bigger studios have an artist dedicated to problem solving the dynamic dress issues and several animators. Unfortunately this character was entirely my responsibility which limited possibilities, but overall I'm pleased with the final results. In the end it taught me the valuable relationship between design for visual appeal and design for technical feasibility. Artists tend to stick with the first and produce interesting stuff, but at the end of the day it has to work and excite at the same time. It was definitely exciting to see this character run around the first time and evolve over production.
