Maria Gehrke shared the workflow behind the Mechanic project, talked about the challenges she faced, and explained why UE5 was chosen instead of Marmoset Toolbag.
My name is Maria Gehrke, I've been doing 3D art for a little over a year now. Before that, I worked as a writer and editor for about six years, but after a severe burnout, I decided to consider another profession and found my passion in character art. I have no art education, and I learned the technical aspects of game character creation through online courses. I continue to develop my skills through YouTube tutorials and with the help of various discord communities. My goal for now is to build a nice portfolio so I can apply for a character artist position.
The Mechanic Project
This project was made for the ArtStation challenge and it’s based on a concept art by Juan Caruso. I was pre-selecting concepts and already had a couple of favorites, but as soon as my gaze fell upon the Mechanic, I instantly knew – this was it. I didn’t even bother to keep searching and immediately got to work.
For me, the challenge was to get the model done in seven weeks. At the very beginning, I made myself a schedule of how much time I will spend on which stage. According to it, I had four weeks for high poly. I think this is the most important stage, and I did not want to rush it.
High poly was fully done in ZBrush. Most of the time was spent on blocking and trying to catch the correct proportions. I wanted my character to be as close to the concept as possible. However, I still had to make a couple of changes in order to save some time and make my life a little bit easier. For example, the most confusing part of the concept was the jacket fold above the waist bag. I decided it would be better to get rid of this fold in order to improve the readability of the model.
In the beginning, I had to deal with the question of how to make clothes. There were options to sculpt it by hand or to simulate it in Marvelous Designer. It was a tough choice. I knew that working in MD would take me a long time, but on the other hand, sculpting folds is not my greatest skill. In the end, I decided to sculpt everything in ZBrush. It was a good exercise, although I am not very happy with the result. I still need a lot of practice.
It was very naive of me to think that it would take me three days to retopologize the model. In fact, I spent seven days as I had to redo some of the parts. On this project, I started retopo in Blender using the RetopoFlow plug-in. Its tools are perfect for the quick creation of the draft low poly. Then I exported it to Maya, where I moved everything into place and finalized the mesh.
I usually do basic unwrap in Maya and then export the low poly to RizomUV. I like this software a lot – it makes straightening UV-shells very easy and automatically packs them. Working in RizomUV was like a much-welcomed respite for me – after a very intense retopology process, I was so pleased to sit back in a chair with a cup of tea and watch how RizomUV does all the work for me.
I textured the model in Substance 3D Painter. Due to the fact that I spent more days on retopology than I expected, the time for texturing had to be reduced. Unfortunately, this had an impact on quality.
All textures were made using alpha maps and mask generators. I deliberately didn’t create tertiary details in ZBrush so I had the opportunity to try various options of detailing in Substance. I am glad that I had a concept and didn’t have to choose the colors myself, it saved me a lot of time. I just had to add more layers of dirt to make the character more believable.
Many experienced artists advise to set up a Marmoset scene from the start and constantly check how the textures look when rendered. This is very good advice, but I arrogantly decided to ignore it because I didn’t want to waste time. In the end, I lost time anyway, because during the rendering process I had to go back to Substance several times and edit the textures.
Initially, I planned to render the model in Marmoset Toolbag with raytracing. I created a scene, set up basic lighting, but the character looked so dull in there that I decided to take the opportunity and try Unreal Engine for the first time. I was worried that I wouldn’t have time to figure out new software, but it turned out to be simpler than I thought.
I didn’t make a pedestal – instead, I used a large cylinder for the floor and background. The idea was to achieve the same look as in the concept. For the cylinder, I created a simple material with concrete textures that I downloaded from Quixel Megascans.
The lighting setup took up most of the time. I left the default Directional Light in the scene and added several additional light sources. There is a Spot Light that illuminates the model from the front and above, and two Rim Lights on the right and on the left. This setup made the character look too flat, so I also added two Point Lights right in front of him to make the shadows a little bit deeper. And the last addition is a rectangular lamp that shines on the cylinder behind the character and gives it a warm brown tint.
I've done about a hundred test renders, tweaking the settings, adjusting the textures in Substance 3D Painter and trying to get the right look. It took me about five days, and sometimes I felt quite frustrated, but I'm glad that I had the patience to push it to a successful result.
The Key to Success
Honestly, I think my project owes its success to an amazing concept art and its author Juan Caruso. It was so much pleasure to work on this adorable character, and I hope it shows through the result.
I am glad that I decided to take part in the challenge and that I made it to the end. It was an important exercise in time management, and now I think I should create schedules and set deadlines for my personal projects too. It really helps to stay organized and keep a nice working pace.
You may find this article interesting