Pedro Varella explained in detail the creation of the Henby sculpt, covering modeling, retopology, texturing, and rendering, and shared his tricks on how to set up scarification without references.
Introduction
Hello there! I'm Pedro. I was born and raised in São Paulo, Brazil. I have been an avid student of character art for games since 2016. Some of the games I have worked on are Predator Hunting Grounds, Diablo Immortal, and Pavlov VR.
My first introduction to game art was during my high school years when I moved to a school that had a Game Design technical course, as well as regular classes. Since then, I have been obsessed with game art, on a journey that has been constantly readjusting its path but never straying from characters.
As years went by, the most important thing has always been to give my very best on each portfolio piece and make it so that I am constantly at the edge of my abilities at every step. Never doing things in a mechanical manner, just repeating what I already know without striving for a deeper understanding of the character art techniques. Mastery of the craft is the aim.
I don't have a game art university degree or anything like that. I couldn't. Brazil did not have any university degrees in game art when I started (2016), and doing them abroad was prohibitively expensive. In addition, the video game industry in my country is still tiny to this day. So, I acquired the skills by finishing portfolio pieces while also doing free courses with the incredible character artists Brazil has given birth to in the past few decades. Some of those teachers were Rafa Souza, Rafael Grassetti, Gilberto Magno, and Igor Catto. Learning with the best is crucial.
The Henby Project
This project started when I decided to open the first two classes of my character art workshop. I wanted to produce a piece of my own while the students did theirs. That way I could use it for some demonstrations and also they could see the progress in realtime as the workshop went by.
So, I found this amazing concept posted by KRAFTON on ArtStation. It was perfect because it had complex drapery for manual sculpting, a lot of hair and beard, anatomy, skin, human face, lots of ropes, and also some aspects of creatures from the horns and the unique skin scarification. All of these were skills I needed to improve or perfect. Furthermore, an important factor that made me choose it is the captivating personality it conveys. The character is not at all generic. You can guess his identity, personality, and history just by looking at the concept art.
References
So, the first step after choosing the concept is building a reference board. While using PureRef, I searched all around the internet for references that could explain to me how to solve specific areas of the character. That includes both photographs and works from other artists, which have something I was not sure how to solve. For instance, let's say I decide I am going to search for references that can help with the high-poly sculpting of the kimono.
So, I go on and look for photographs of old and new kimonos, close and wide shots. Also shots of high-poly models with torn fabric, similar kimonos, similar fold sculpting, etc. I do the same for all parts of the character (hands, face, legs, accessories, etc) and repeat the search focusing on the subsequent steps of the pipeline, especially the texturing phase. In addition to this, I gather some images of works from other artists I deeply admire. Those will serve me as references and quality goals, even though I am aware that I will probably fail to reach their level. They are much more experienced than me, and that’s just a fact of life. They will be on a different level. Still, I must aim high and do my best to put myself on my limits.
The reference board is never finished. As I work on the project new questions arise, and I look for new images that can answer them.
The Face
I already had some references for people who have similar Asian features and facial expressions. So I imported a simple human base mesh into ZBrush, smoothed it away (to make sure I don’t keep myself attached to the base mesh forms), and started sculpting. On my Instagram, there are some videos of the sculpting of the face. My process of sculpting is strongly based on traditional techniques and has no complex software usage. It's about a solid understanding of anatomy, years of practice in sculpting the human figure, and using the basic ZBrush brushes (Standard, Dam Standard, Clay Buildup, Smooth, Move, and Pinch). The base mesh topology is not a concern at this phase. I am free to stretch and distort it as much as needed.
I do not work linearly on the model. I did not finish the face before moving on to other areas of the model. Quite the opposite. Once I worked for some hours on the face, I went to another area, then another, then back to the face, and so on. That creates a nice flow of work and allows my eyes to rest. Every time I get back to the face, I come with fresh eyes and see new things to improve.
However, this character's face was especially challenging to me. His facial scarification is not only something I have never seen in any other character, but it also does not exist in real life. So, I saw myself in a place where I could not find any useful references for this. The only things I had were references of third-degree skin burns and African scarification, which were not quite what the character has. But I had to work with what I had, so I spent a while sculpting, erasing, resculpting, and exploring, until I started to understand what visual direction I should take his skin facial scars to. While sculpting, the vision solidified itself more and more and I could sculpt the rest of the face scarification with more understanding.
The skin detailing was made using 3D Scan Store's amazing animation-ready models. I chose an Asian one and used the ZWrap plug-in inside ZBrush to snap the scan model over my own sculpt. Alongside the use of layers, the process creates a model that has the forms of my sculpture and the very high-resolution skin detail from the animation-ready scan. That also allows me to use their color, roughness, and specular textures later on. But do not be mistaken, I didn't use this as the final low poly. This process has its only purpose in creating a highly detailed face and body for my character that was used as the high poly model in the baking phase. My final low-poly was created manually in Maya at the retopology phase. I should also mention that by the time I started working on the skin detailing, the entire high-poly model of the character was ready for detailing as well.
For the horns, it’s pretty simple. I inserted some spheres into place and started sculpting the primary shapes, focusing on silhouettes. Then sculpting the secondary forms using the Clay, Clay Buildup, Standard, and Dam Standard brushes.
The Clothing and Accessories
I take great pleasure in sculpting folds by hand. I understand the purpose of Marvelous Designer in a production environment and I use it constantly at work. But there’s nothing I love more than sculpting. As I see it, a Character Artist is, before anything else, a sculptor. Marvelous Designer takes much of that joy away from me by turning a sculptor into a couturier. That is even more relevant considering that the concept has some unique drapery on the kimono, waist, and shins which I wanted to capture.
None of this means that it couldn’t be done using MD and then refining in ZBrush. It definitely could. Marvelous Designer is an amazing software and it’s extremely useful in production. But I personally love sculpting folds by hand, so I did.
Basically, the workflow for the entire clothing and accessories is the following.
- Create a geometry for the piece using a sphere or extracting from the body.
- Sculpt a rough sketch that prioritizes the primary forms and the silhouette. This is when you add all the objects that compose the character. Everything should be roughly in place.
- Then sculpt secondary forms while flowing from one piece to another and not spending too much time on one single piece. Refrain at all costs from detailing anything. This is the stage where the character starts to look a bit better and gain personality.
- Once most of the pieces are sculpted, you go on to each SubTool, separating objects that were sculpted using the same DynaMesh geometry and refining each object at a time. Cleaning surface wobbliness that may have happened during the secondary stage is common in this third stage. This is when the character starts to look final as you refine each object.
- One challenge for me during the sculpting of the Henby high-poly was making sure the sculpture did not look too clean. It had to look rough, torn, dirty, and handmade. Cleanliness is a natural tendency in any 3D work and I focused a good amount of my energy into solving this problem.
With the exception of the skin, I do not fully detail the high-poly model to the point of adding high-frequency details (tiny and repetitive surface information). That is added on the texturing or shader phases using tileable textures. So as you can see in the screenshot below, the high-poly model is completed at the tertiary forms. For instance, fabric tears, stitches, and smaller fabric wrinkles were sculpted, but not textile fiber textures.
Retopology
Before starting the retopology, it is crucial to prepare the decimated high-poly models. The process is simple, but I have never seen anyone talk about this on the internet.
- In ZBrush, delete all SubTools that will not be used during retopology (ropes, placeholders, etc). Leave in the ZBrush scene only what is going to be used as a snapping surface to create the low-poly model over. Also, objects that are repeated or symmetrical can be deleted and left with only one. For example, Henby’s shins are symmetrical. So I deleted one of his shins and only made the retopology and UVs of the left one.
- Look for geometries that will be the same topology on the low-poly and merge into one SubTool. Turn on Backface Masking and, using the Move brush, push the geometry inwards where there are holes. Then apply a high-resolution DynaMesh (if you come from 3D printing you probably are familiar with this process). This will prevent the high-poly model from having internal geometries during retopology. It should have one single shell for the retopology to snap over. In other words, make it "watertight".
- Decimate everything to a level that allows for proper retopology and is light enough for Maya to handle. I usually decimate it to around 1% or less.
As I finished the high-poly preparation, I exported all SubTools to an FBX file and imported it into Maya. Then, I selected one decimated high-poly, turned it into a live object, turned on Maya’s “Modeling Toolkit” and using the Quad Draw tool I created the polygons that are going to be our low-poly model.
There is a specific logic to the retopology technique I use in most of my characters. Allow me to explain. The idea is to have a low-poly model that is light, animation-friendly, and also completely retains the high-poly forms and silhouette. All at once.
- The first step is to create a very light cylindrical topology which is going to be what makes it animation-ready. This is easily destroyed during the next step process, so you must be attentive not to. Then you can probably subdivide it once or twice depending on the aimed final polycount.
- Secondly, switch to the Multi-Cut tool from the Modelling Toolkit and start adding cuts wherever necessary completely disregarding the topology. You shouldn’t delete or merge vertices created on the first step.
- Thirdly, add cuts to get rid of polygons with more than 4 vertices.
This process will result in a topology that is layered in a sense. First, you have the cylindrical topology that is great for animation. On top of that, cuts are added. These are the ones that retain the high-poly shapes and silhouette. They also make it unnecessary to create super complex topology flows to retain the shapes.
I know it may look a bit messy, but if you do it right, it will work just great in-game. Some people may think this is wrong because the model will end up filled with triangles. This is a vice inherited from the VFX industry to prohibit triangles in an animated model. That is still true for that industry, where the models are later subdivided and triangles create small artifacts on the surface. But that is just not the case for games. low-poly models are never subdivided, not for characters at least. So as long as we create an animation-ready topology right from the beginning and retain it until the end, we can use as many triangles as we want.
UV Mapping
The UVs were also created manually. Everything was made in Maya. So after adding all the cuts and unfolding everything, I did the process of straightening as many UV islands as I could. In the GIF below you can see how it’s done. Select the sides, straighten them, and unwrap them horizontally. Then do the same for the top and bottom vertices and unwrap it vertically.
I pay attention not to do this on islands that are too circular or organic in shape. I did this wherever I could but not on everything. The layout was basically done manually, placing one island at a time. No automatic layout can do a better job than a competent manual placement. The only guideline is to start with the largest islands and then go on placing the smaller ones. I did not worry about island orientation. It doesn’t matter. The direction of the tileable textures can be adjusted in the texturing phase.
Texturing
The skin texturing was done starting with the color texture provided in the Animation Ready scan. I took the color texture, applied it to my high-poly model in Zbrush (which if you remember, has the scan's UVs), converted it into polypaint, and then cleaned the texture on areas such as beard, eyebrows, and also spots and freckles that I didn’t want. For that, I used the “Standard” brush with “Zadd'' turned off and changed the stroke mode to “Color Spray”. That is very important to bring a nice color variation while painting the skin.
After cleaning the colors in ZBrush I converted the polypaint into texture and exported. Then I imported that texture into Substance 3D Painter, changed the colors to a red tone using an HSL filter, and painted more colors and details to correctly match the visuals of the concept art.
For the rest of the character, the textures were painted in Substance 3D Painter from scratch. The focus was to paint textures that have a lot of variation both on the Albedo and Roughness. Avoid at all costs cleanliness.
After applying the base materials to each object with an estimated color and roughness, I brightened the peaks and darkened the valleys of the surfaces using the baked curvature texture. Next is the color variation itself. That came by manually painting very subtle purples, blues, yellows, and oranges in some areas, adding dirt using masks, and applying a color noise in overlay blending mode over some materials.
Hair Cards
The hair textures were created using a mix of simulations using ZBrush’s fiber mesh and manually painting the strands for the cards in Photoshop. I basically created one card at a time. Made one alpha texture, created the card mesh, and already started manually placing it in ZBrush using the Gizmo “Bend” deformation.
Only after I was done with those first primary cards, I created the next one. This is so that I don’t have to predict how dozens of cards have to look in advance before even starting to make the hair. I believe it is much more organic to create and place one card at a time as you feel the need for it. That way you can precisely decide how the next card must look based on what you did before.
But the overall logic of the hair cards is:
- Primary cards (very thick and filled with hairs; supposed to make the overall shape and volume).
- Secondary cards (less thick, with some open areas and already with color and roughness variation; supposed to start adding subsequent clumps and some variation but follow the primary direction of the hair).
- Tertiary cards (has visible clumping in the cards and most of its areas are transparent; supposed to break the surface of the secondary; still following the basic flow of the primary and secondary).
- Breakup cards (thin clumps that will be free to not follow the flow of the previous cards and add a finer look to the hair).
- Flyaway cards (cards with a small amount or even individual hair strands that are completely independent; as the name suggests, they are supposed to be crazy).
- Transition cards (short cards that are supposed to help with a natural transition on the hairline).
Rendering
The rendering was fully done in Unreal Engine 5 and composed in Photoshop. Once I had the character's shaders done, with all the textures and adjusted values in the parameters, then it was time for setting a lighting scene.
Henby's was pretty simple. The usual and great 3-point lighting. One key light coming from above, quite big for soft shadows and stronger than any other light source in the scene. One Skylight with a custom HDRI and low intensity, which works as a Fill Light. Two rim lights coming from the right and the left, to create that light that contours the character. And finally, one last light is used to add that highlight to the eyes. It's positioned and directed right at his eyes and is very subtle.
It is important to set up a cinematic camera and adjust its parameters while viewing through it. That way you can actually see what is going to come out from the renders using the Level Sequence. The viewport itself does not represent the final look.
Some of the post-effects used in the camera's settings were Vignette, Film Grain, Depth of Field and Aperture, Chromatic Aberration, and Exposure. No secret here, just tweak them until you find the visuals you are looking for.
The most important thing when creating a 3-point lighting setup for a character is to have a hierarchy between the lights. The strongest and most important one must be the key light. No other one should be close to its intensity (sometimes the rim lights can be an exception). That is one of the most common mistakes I see from students. Having all light sources equally strong completely kills all shadows.
Another tip is to pay attention to the size of the light source relative to the character. The larger the light, the softer the shadows. But the distance between the character and the light is also relevant. The sun is a massive light source, but it's also super far away, which makes it small in the sky. So from the sun, we get harsh shadows, not soft, despite its size.
The Journey
This character was done with very limited free time for personal reasons. So I had to exercise discipline and work on it every single day for about 1 hour. The pace of improvement was very slow, but steady and constantly challenging, which is the most important thing. However, the total amount of time dedicated to it was about 6 weeks, which is very much reasonable for its complexity.
Being chosen because it had technical challenges all around, the Henby character was demanding on a daily basis. But the most challenging parts were the hair (which is the step of the pipeline that I had the least experience with compared to the rest), the skin detailing and texturing because I was practicing brand new techniques, and the UE5 shader and rendering, which were also quite new to me. I was much more used to rendering characters in Marmoset.
My approach is to push myself to the limit and try to do something better than ever before at every step of the process, even if I'm doing something I've done a thousand times. Always keep myself at the edge of my abilities, which is tiresome, but exciting and worthwhile.
Final Thoughts
My most important advice for a beginning Character Artist is to finish portfolio pieces. I see way too many artists spending years making WIPs and never finishing anything. In my understanding, that's a very effective way to get exhausted from your own dream.
If you do not finish personal projects you will have nothing better to add to your portfolio. The probable consequence of that is you'll eventually grow tired of waiting for better jobs to come your way. Being stagnated professionally and always working for bad clients while being poorly paid.
So, finishing personal projects is not an option. If you want to keep loving what you do and doing what you love while getting well paid and growing as a professional, you must finish some personal projects regularly.
Also, when doing your projects give your heart and soul. Keep yourself at the edge. Do not do your projects as if you were washing dishes, mechanically and mindlessly. This is your craft and your dream. Always do your best and be patient. Great rewards come as a result of your true effort.
I hope this was of some use to you. Thank you for reading and happy studying!
Pedro Varella, Character Artist
Interview conducted by Theodore McKenzie
Keep reading
You may find these articles interesting