CGMA - Next Gen Character Creation Mentorship

2021 | Winter Registration now Open!

Registration for our 2021 Winter Term is now open!

Next Gen Character Creation Mentorship

This advanced 8-week course has been designed to learn in-depth techniques for modelling, texturing and rendering a cutting edge real-time character.

Course overview Course overview

Course Overview

Create AAA level game characters

This course has been designed to learn in-depth techniques for modeling, texturing and rendering a cutting edge real-time character. The class will work similarly to a live mentorship, as students approach the creation of AAA game characters for their portfolios. Students should expect to cover head and hair, costume elements, low poly UV’s and processing required to get the asset real-time ready, and finishing with material and texture creation to set up the final model in engine with final images. Please note that this is an advanced class for students to push existing character(s) and portfolio pieces.  NOTE: This is NOT an introductory course. Recommended prereq: Character Creation for Games.  Students must have intermediate to advanced knowledge of anatomy, Zbrush, and Maya. If unsure about skill level, send your portfolio to Academic Advising. 

More  
Less  

Course Format:   All live group sessions
Lecture Type:   Live group sessions (recording available afterwards)
Feedback:   Live group feedback (recording available)
Duration:   8 weeks
Assignment:   Deadlines each week
Q&A:   Once a week
Materials:   Computer, Zbrush, Maya or 3DS Max, Painting software such as Substance Painter, Mari or Photoshop, Marmoset Toolbag.
Skills level:   Advanced
Prerequisites:   Must have intermediate to advanced knowledge of anatomy, Zbrush and Maya or Max. Course pre-reqs recommended: Character Creation for Games, Cloth Creation and Simulation for Real-Time and Hair Creation for Games

Next Gen Character Creation Mentorship WHAT YOU’LL LEARN

What you'll learn

The more you know, the better.

expand all
collapse all
Gathering Reference | Creating to scale blockouts | Primary Forms | Real world Scale | How to start the model | Reasons to block out proportions
Finish Primary forms of the face | Work on Secondary forms of the face | Developing secondary anatomy of the face | Adding Character without overt expression | Using references | Adding Asymmetry
Methods of adding pore details | Additional Skin Breakup | Adding hand sculpted details | Balance of primary, secondary and tertiary details
Finding patterns | MD tool overview and workflow | Garment Creation | Tips and Tricks | From MD to Zbrush Workflow
Detailing MD simulation in zbrush | Discuss detailing in model vs texturing
Discuss edge flow for deformation | Learn retopology tools within Maya | Discuss UV layout workflow in Maya, where to optimize UV and why | Learn UV tools within Maya
Learn Xgen Basics which can be used for pre-rendered or real-time hair creation
Lighting and Composition | Final Render and Portfolio Images | Final Render setting | Final Camera settings and effects | Photoshop to present best possible images | What to show in Portfolio
Instructor

Bringing out the best in talent

Adam is a Senior 3D Character Artist currently living in Cary, North Carolina, working at Epic Games. He has had the privilege of working on The Order 1886 as Lead Character Artist, as well as on Uncharted 4: A Thief's End--two of the most visually stunning games of this generation. At Epic Adam is a part of the special projects team where he contributes to next gen character development in UE4 as was seen at Siggraph 2017 with the Meet Mike demo.

Student interviews

COURSE BEGINS

April 17th!

Winter TERM Registration

Oct 26, 2020 - Feb 1, 2021

Only

$899

COURSE BEGINS

April 17th!

Pricing & Schedule

Even though our courses are the most affordable for the quality of education.

These Finance Options allow you to focus on your goals instead of the barriers that keep you from reaching them.

Employer Reimbursement

Animation Guild CSATTF

Payment Plan

Companies that hire our students

  • Naughty Dog
  • Luma Pictures
  • Google
  • EA Games
  • DreamWorks Animation
  • Blizzard Entertainment

Speak to an advisor

Need guidance or course recommendations? Let us help!

Have you taken a course with us before?

Show us your skills

Not sure if you have the skills, or are you proving you do? Show us.

Have you taken a course with us before?

Shaman Production

Interview with Alejandro Olmo

Alejandro Olmo talked about his Norse Shaman character made within CGMA course Next Gen Character Creation with Adam Skutt. Software used: ZBrush, SP, Toolbag, Mari, Marvelous Designer, Knald and more (check the full list on ArtStation).

Shaman Production

Introduction

My name is Alejandro Olmo and I am from Puerto Rico. I am a character artist for movies and video games. Currently, I am a freelancer doing voluntary work for a video game called Galaxy in Turmoil for Frontwire Studios. I have been blessed to have the opportunity to work as a concept artist at the early stages of the upcoming Sony Pictures film Venom, and as a 3D modeler for the upcoming motion pictures Replicas. My background also includes smaller projects for 3D printing, 3D modeling for video games characters and as a Professor and 3D Character Artist in the Creative Department for a local university.

CGMA Course Goals

My objective in taking the Next Gen Character Creation class at CGMA was to learn the correct pipeline in the industry for a real-time character. The process was completely new to me because all I have worked on before was concepts and basic 3D models. I wanted to understand my weaknesses and learn how to focus on a specific project and make it look good. It is important for me to practice every day, learn something new, keep pushing myself to get better because my main goal is to get a full-time job in the industry.

Start of the Character

In terms of sculpting each character presents its own challenges. Anatomy is one of the skills I try to keep learning. One of the key elements that allow you to get more value taking a CGMA course is receiving the feedback of your mentor, in my case was Adam Skutt, which by the way is one of my favorite artists. I am not the best at anatomy and his feedback helped me do a lot of drastic changes to the character. There is still much to learn and every day comes with the opportunity to keep learning. The best thing you can do to make a character look good is to use references, without them it is more difficult to achieve the objective.

I always start all of my characters by gathering references. Once I have what I was looking for I start the sculpting process in ZBrush.

Reference & Story

The first step of the process was to look for references as I said above. The details of an organic model like this one are important, it requires time and a clear visualization to make it pop. The process of creating the character, accessories, and clothes go hand in hand. I wanted the character to tell a story, allow the viewers to imagine where he lives, his behavior, his traditions, etc.

My biggest inspiration to create The Norse Shaman came from two of my favorite games: The God of War and HellBlade Senua’s Sacrifice. I wanted to create a character that could easily fit into one of those worlds. I didn’t want to create a heroic character. The main goal was to create a character that makes you wonder: is he good or evil, does he have powers, can I trust him? Something a little more complex. That is why I decided to create an old man, a little deteriorated with tattoos and either little or no expression.

Skin Shader

Creating a real-time character and making sure that shaders look good is new to me. To be honest, it is all new to me. I am still working on the character, looking for ways to improve it, to make it look more realistic. I took the liberty of contacting some artists that are already in the industry and asked for feedback. One of the things they all agreed on was that I needed to re-work the skin shaders. I used Mari for the color base and Substance Painter for the roughness and dirt on the skin. For the translucency, I did a bake in Knald and applied it to Toolbag working with some parameters. I also tried to maintain a simple illumination inside Toolbag, basically, a 3-point lighting and a very soft HDR fill light.

Here is a before and after of the skin shader:

     

 

etailing

To create the details inside ZBrush I use the standard, dam standard, and clay brushes, then I exported a decimated model to Mari and using textures of Texturing XYZ I generated the displacement map. Next step would be exporting it back to ZBrush to have the basic and micro details to do a bake in Knald afterward. I added a noise texture in the detail normal map channel inside Toolbag to have more defined pores and light breakup.

Toolbag Renders

     

 

 

Feedback: Challenges

One of the biggest challenges I’ve encountered was to understand the full pipeline and be able to implement it on the character. I also had to learn to use new programs like Mari, Substance Painter, Marvelous Designer, and Marmoset Toolbag. Every character has its own challenges but I could say that anatomy, proportion, and materials are the common ones. I am still working with this character, practicing and learning new techniques every day.

Feedback: Lessons Learned

The biggest lesson learned is the dedication and time that is required to create a high-quality character. I had to learn how to be exposed to constant constructive criticism and not to be afraid to receive feedback from other artists in the industry. Honestly, it helped me grow and it has kept the motivation to keep wanting to learn more, be a better artist and fight more to achieve my dreams.

  

 

Alejandro Olmo, Character Artist

Interview conducted by Kirill Tokarev

Real-Time Character Creation

Interview with Bailey Wheatland

Bailey Wheatland prepared a great breakdown of his real-time character Amasol: building a female figure and facial features precisely, creating voluminous hair, experimenting with shadows in UE4, working on the corset material and much more. 

 

 

Introduction

Hello, my name is Bailey Wheatland. I am a character artist and a recent graduate from the Savannah College of Art and Design (SCAD). I currently reside in the small town of Streator, Illinois where I have lived for most of my life. As a child, I always had a massive interest in video games. Some of my fondest memories include me and my family having Crash Team Racing tournaments at my aunt and uncle’s house. This formative gaming experience is what led me to love them as deeply as I do now.

For as long as I can remember I have always been referred to as the “art kid”. I would make my own comics, create new dinosaurs, and write short stories just to keep training up my creative muscles. I took a 2-year break from art during middle school and it wasn’t until then that I realized that my love of games and my passion for art could coexist. This realization is what led me back to creating art. It became this new thing of knowing exactly what I was going to do with my life.

This journey into 3D art began around the time I was 13 when I started learning Blender.

I realized early on that my main interest was in creating characters and creatures. There was something that appealed to me about organic forms as opposed to hard surface ones. I used to model all my organics in Blender with a trackpad on my laptop at the time.

 

I remember modeling the Thanator from Avatar with that setup. that was the last organic object I made from scratch in Blender, primarily because around this time I picked up a copy of 3D world magazine that featured an awesome looking dinosaur on the cover (what child wouldn’t gravitate towards that?) It was based on a dinosaur concept Greg Broadmore had done for the magazine ImagineFX (which I also had a copy of).

During school, I was a part of many projects that gave me a ton of experience in regards to pipeline development and collaboration. These projects included two animated short films, one live action sci-fi film, three video game projects, and a few board game projects. A majority of these projects were created alongside my coursework, so I was always very busy at SCAD.

Amasol: Start of the Project

Thank you! I had originally selected this character for Adam Skutt’s class Next Gen Character Creation at the Computer Graphics Master Academy (CGMA). The goal of the class was to create a realistic human for a real-time environment. This was our only guidelines in the class so I came up with a few self-imposed guidelines of my own.

  • Firstly, I wanted to create a female character since my portfolio lacked that area of character art.
  • Secondly, I wanted to create a character with pretty intricate hard surface elements to contrast with the innate organic nature of the human figure.
  • My final requirement for the project was to create a character that was just simply beautiful.

I spent 2-3 days searching all over the internet for a character concept before the first class at CGMA. For a while, I couldn’t find a concept with the qualities I was looking for. When I came across an astounding piece by Shin Tae Sub for Legend of the Cryptids, I knew in an instant that this was the character I wanted to tackle for the class. I saw the beauty in the image, but I also saw all of the challenges that came with it. The latter was my primary reason for choosing this character.

Since the class was focused on realism, I knew I wanted to change a few things in the concept to bring it a bit out of fantasy and more into reality, which I will touch on later.

Sculpt of the Body: Getting the Figure Right

Before I started this project I was working on another character project of mine that featured Mediterranean figures. And it happened that the topology was already good to go, so I just used that base mesh for Amasol. It may have been easier to have started from scratch, but the challenge of taking a very muscular male and turning him into a curvy sun goddess seemed like a fun idea at the time. It was also very educational in the long run because you make these neat associations between what makes a character look masculine and what makes them look feminine.

 

Over the entire project, I also looked at very many reference images of women. 3d.sk was an excellent website for it, plus Pinterest helped with general anatomy breakdowns. After finding a few women that had a similar body type to what I had in mind, I began working on the figure. On 3d.sk they give you so many different perspectives of the same model. Once I found one or two models, I would just keep clicking through the images as I revolved around my sculpt so as to match the view I was referencing.

Another great reference I had was 3D scan data of a female model. I took an anatomy class with Ryan Kingslien a year ago and with that class, I got a bunch of scans of the same model in different poses. It was so beneficial to see how those forms looked in ZBrush with the same material as my model. The biggest question for me about sculpting realistic humans was how much or how little to push forms and details to get that proper design hierarchy seen in the real world. Luckily with 3D scans, that question has been answered. That solution is also related to what skin shader you will be using since some of them diminish the detail more than others.

There was an issue when I began to block out the primary proportions of the hair and the body. In the illustration, her legs were too long if you draw beyond the image. I quickly found out that there was a trope of fantastical female characters having incredibly long legs. In 3D this created a character that looked a bit disproportionate and didn’t capture the elegance that was seen in the illustration.

Since I needed to design beyond the boundaries of the image, I came up with a concept that would resolve these issues. I wanted to extend the legs through realistic means so my solution was to put her in heels. In this way, I am paying homage to the fantasy proportions of the original work while putting a more realistic spin on it.  

The scale of the hair also played massively into the elegance of the figure. If the hair was too big, she looked very childish and short. If the hair was too small her presence wasn’t as strong and deviated too much from the concept. This took a lot of tweaking after the texturing process took place to get right.

 

In terms of pushing realism in the figure, I would say breaking up the manufactured look to the anatomy is the biggest thing. This is caused by an issue of the primary forms being lifeless due to a lack of fluid lines in the structure. Once you lay down your initial primary forms, you should give them a little bit more character and randomize some of those neat straight lines to create a more dynamic figure.

I normally do this when I reach the secondary and tertiary stages of my sculpture. Another personal trick that helps me is getting to the skin detail a bit earlier in the process once my primaries and secondaries have been established.

For this project, I used a combination of projecting TexturingXYZ textures with Mari and employing noise passes, alphas, and hand sculpting in ZBrush in order to create the skin texture. Once I go through this process of adding “flesh”, I go back a subdivision level or two and continue working on my tertiary and some of my secondary forms, along with the occasional tweak to a primary as long as it isn’t such a massive change that it would distort the skin texture.

This newly added noise to the polygonal surface of the sculpt fools my mind into thinking I’m sculpting a real human. I will use a combination of the Inflate brush set to an intensity of 5, the Standard brush set to 10, and the Clay Build Up brush with no alpha set to a 6. With the inflate and standard brushes I honestly feel like I am injecting fat underneath the skin. In my case, this led to a very organic result due to the build-up of subtle form creating a surface that reads as flesh and not as rigid forms.

 

Once the game resolution model is in the engine, I will further this detail by adding a micro normal map which can be tiled and really helps sell the illusion of skin for up close shots.

Facial Features

There were a few supermodels and actresses who this concept reminded me of so I put a reference sheet together of features that I thought these models shared with Shin Tae Sub’s work. The models I chose were Adrianna Lima, Ashley Greene, and Stella Maxwell.

 

Once I had my references, my approach to the head was very similar to the process I described with the body. I started with the primaries, then added some break up with the secondary forms, added a few tertiary forms, and then started projecting the XYZ textures in Mari. I then imported the secondary, tertiary, and micro grayscale textures into ZBrush as displacement maps where I set each map to a different layer. This allowed me to get the right blend between the different levels of detail. I then continued to sculpt the head, placing additional form breakup around the fattier areas and adding a bit more asymmetry to the face.

 

Once the face was done, I added peach fuzz to the sculpt with fibermesh in ZBrush. I then groomed it with the snake hook brush and zproject brush. Peach fuzz and body hair have specific flows that they abide by so grooming the hair in a specific way is pretty important.

I then baked the peach fuzz in a separate normal map along with vertex color that would act as a mask so I could then control the fuzz normal in either Photoshop or UE4.

 

For the albedo map, I used Substance Painter and followed the method laid out by Magdalena Dadela which can be checked below. This method includes using a smart material as a base and then adding the different tonal zones to the face along with mottling and the other specifics to the skin. I highly recommend watching that video since it will go into greater detail than I will here.

In addition to the methods discussed in the video, I also added a peach fuzz layer to give the strands a bit of color in the albedo to fake the skin reflections that you would see on actual peach fuzz. The vertex color I baked off earlier acted as a mask for my fill layer in Substance Painter.

Once I had the initial texture done, it was just a matter of importing it back into the engine to see how things were looking. This was tested in a look development map I had created, and in the map we used in the video.

Working on Hair

Before starting on the hair I gathered some references for the flow of the hairstyle. A big challenge for this project was creating a wavy style along with posing it in an animated way to give it the sense of floating in the air.

The workflow I chose was a blend between Tom Parker’s approach and Adam Skutt’s approach. I originally created my own insert mesh brush divided into 3 segments in Blender. You have one part for the root, the middle part that can be continually repeated to add length, and then the ends. Once I got my curve brush set up in ZBrush with the mesh, it was just a matter of drawing each strand out and placing it by hand.

I had the idea of doing a simulation in Marvelous Designer with the hair strands at one point, but since my UVs aren’t useable in that program, there was no way I could pin the strands at the root to get a good simulation. So my other solution was just to mask the roots in ZBrush and then use some of the deform modifiers in there to help me out a bit with getting that light quality to the hair. I was able to quickly mask the roots by flattening the model with the uv master, box masking them, and then unflattening. For some reason, this only worked the one time and all other times I tried this the masking would always get lost between unflattening and flattening. Fortunately, I was able to just use the mask pen and mask the roots fairly well on the model without that additional step.

During this initial process of laying out the hair, I quickly drew a few texture strands for it just so I could get an idea of how much area I would be covering once the planes were textured.

 

These were just quick-and-dirty sketches that I would then replace with hair textures generated with Xgen in Maya.

I only rendered out an alpha and AO map for the hair. Unreal has a powerful hair shader that requires a few additional maps so I went ahead and made those for faster look development iteration.

These textures included a root to tip, hair ID, and pixel density offset (PDO) maps. I was able to get away with just using constant vector 3 (CV3) nodes to create the hair color. For those unfamiliar with programming or Unreal, this is essentially a node that just offers me any singular color I want, much like how you select colors in Photoshop. The 3 vectors relate to the R, G, and B channels. With the ID map, I was able to get color variations between each strand of hair as well as variety in the values of the hair.

Two other textures I created came from altering the AO which was the normal and the flow maps. The flow map helped blend the planes together so they didn’t appear so segregated from one another. The normal map allowed me to add extra depth to the individual strands which helps a lot for close-up shots.

From then on it was just back and forth between ZBrush and Maya as I continued to shape and place the hair all while adjusting the UVs by hand for the further breakup between the layers of hair. I wish the process would have been a bit more streamlined, but I’m happy with the result I was able to achieve in a couple of weeks time.

 

I was able to push Epic’s hair shader a bit more by changing the shading model from hair to subsurface profile. This allowed me to not only achieve a softer look to the hair but also to control the extinction scale of the effect which allowed some very nice light to pass through the strands in combination with the PDO map.        

    

Texturing & Colors

The textures for the clothing were very simple. The entire process took me around 20 minutes from start to finish in Substance Painter. With this project, I actually did shader work before texture work which was an absolute reversal of how I normally approach character art. If this were in Marmoset I would have done textures first, but there is so much more you can do with the material editor in UE4.

I did most of my legwork with the sculpt already so when it came time to texture, I used a combination of texture bakes as masks in Substance Painter. I had curvature, AO, cavity, and world position maps for the metal parts of the outfit. By using one of the metal smart materials and a little bit of tweaking and hand painting, I was able to get a good result in a very short amount of time.

I normally multiply the base texture with a flat color (CV3 node) in UE4 so when I made a material instance from the master material, I could then alter the color instantly to match the colors in the concept more effectively.

As I got further along in the process, I did alter the color scheme a bit from the illustration, because the lighting of the scene called for a different approach towards how the colors interacted with one another. For instance, the value of the corset changed a lot over the process. Sometimes a higher value looked better so it could stand out from the chest piece and other times a slightly darker value really made it stand out from the skin. It just took time before I settled on a happy compromise that was everything I was looking for.

Corset Material

The corset is the material I’m most proud of. This was the most complex material I put together on the project, and while it took some tweaking, it came out better than I had envisioned. I knew the general approach to get the material done and that’s why I wanted to have my final character done in UE4 and not Marmoset. You can create beautiful images in both, but the flexibility of the material editor in UE4 is what made this solution a no-brainer to me.

The goal of the corset was to make it a two-layered material. It would feature a cotton twill material at its base and a silk sheer material on the top. I chose to use the clear coat shader as a starting point for the material since it utilizes 2 normal and 2 roughness map slots.

I knew I was going to be utilizing two normal maps so I considered that when creating the high poly version of the corset. I started by creating the base in Marvelous Designer and referenced a sewing pattern of a corset I had found online. Once I got the base done I then brought it into ZBrush to further refine it. After the base sculpt was finished I then sculpted the sheer material on another layer. This allowed me to toggle that sheer structure on and off so when I got to baking the textures, I could easily export two different high poly versions of the corset.       

   

 

The sheer texture was taken from a picture I took at my local Hobby Lobby that had the exact structure I was looking for. I created an alpha texture from that image and that is what I used to begin constructing the top layer.

When making the shader for the corset, I had originally wanted to use the cloth shading model, but it didn’t offer the layered effect that I wanted. I noticed that one of the cloth-specific texture slots called fuzz replaced the emissive texture slot. So in the clear coat shading model, I treated the emissive slot like the fuzz and I was able to get the same effect seen in the cloth shader.

The biggest challenge in the material was to get the opacity fine-tuned in a way that allowed enough translucency on the border of the model as well as having some opacity shifts within those borders. Sheer is a very translucent material so it was paramount that I got the effect right.

My solution for the effect involved using a combination of fresnel nodes for the opacity mask, emissives, and albedo. Most of this was driven by a thickness map I had baked out of Marmoset that contained the sheer corset high poly information.

Corset before texture work in Substance:

      

Once I had all my textures I started to plug them into my shader and, with material instances, I was able to tweak the values of the fresnel masks to my liking.

Presentation: VFX, Animation, Lighting

The initial idea to do a short video came about when speaking with my good friend Antonio Gil, who handled the video editing, VFX, and environment for the project. We got to talking a little more about the collaboration and we both decided we wanted to see it through. I had already done a short video for my previous character project so I figured this would keep in-theme with that showcase method.

With Antonio still being at SCAD, this project was what he pitched in one of his studio classes. From that moment on, we were on a 10-week deadline. By the time his quarter started, I was already in a good place with the sculpt. I probably spent one more week tweaking it and then began the retopo process so we could get the game ready asset underway.

Throughout the quarter there were suggestions from Antonio’s professor that initially had me stressed out because of the quality expected and the time we had. One of the suggestions was to employ a good bit of animation which included blinking, breathing, and hair movement. I am not an animator so my mind was already reeling on how I was going to get this done.

Luckily I remembered I had done a degree of the limited animation we wanted with a different project two years prior. My solution was to use blend shapes that I could call via a timeline in Unreal Engine 4. This solution allowed me to animate directly within Unreal.

The drawback to this method is that you can’t currently track morph targets with Sequencer in UE4, so the solution I had was to call the sequence at the beginning of hitting play via the level blueprint. This allowed me enough flexibility to get the timing down for our shots.

Due to time, posing the character all in ZBrush would have been tricky so my friend John Aaron Lohrey provided me with a rig using Human IK in Maya. This allowed me to quickly get the main pose for the character. John also provided me with eye and mouth controls that could be used as well. At the end of the process, I had 7 blendshapes I could animate with timelines in UE4. I had an idle pose, head turn, blink, and breathe, as well as eye, finger, and mouth moves.

The lighting was a shared responsibility between both myself and Antonio. Antonio did the initial lighting for the environment. I then matched that lighting on the character with area lights provided by the engine.

For the final image, I ended up redoing a good portion of lighting due to our previous map in UE4 crashing when trying to open it. Fortunately, this was after the quarter and after the final turn in for the video. I had duplicated the map, but in that version, the lighting needed to be rebuilt. Every time I tried to rebuild it, the project crashed so that is why I needed to alter the lighting in a way that roughly matched what we had for the video.

When it came to the final beauty renders, I edited one of Antonio’s particles to give me a cool circular shape around the energy sphere. The trail was just created in photoshop for an added kick to the image

Feedback: Lessons Learned at CGMA, Experiments, Reflexion

This project took 3.5 months to complete and it taught me so much in terms of patience and iteration. I learned a couple of things in the CGMA class, but most of what you are seeing is a culmination of my education over the past 4 years. Having just graduated, I now have the time to apply everything I have learned to my projects in a much more effective way than in years previous.

I would do so much outside of school whether that be working on collaborations or participating in other course work. For this reason, my mind was always in a state of “go, go, go” so I didn’t get a chance to think on what I learned as much as I would have liked. I understood the content, but I still had so much to figure out in terms of applying it.

Adam’s CGMA class was my one educational obligation this past summer so I committed to that class with my undivided attention. I didn’t learn much new information since I already had access to Adam’s course through recordings on Game Art Institute. I applied for the course primarily to get Adam’s feedback on my work. The course to me was kind of a giant test of my current knowledge and I used Adam’s feedback as a point of validation for said knowledge.

If I were to give a few insights into the specific lessons learned with this project I would say two big ones are the importance of lighting in displaying character details and the patience needed to craft a project of quality.

In terms of lighting, there is a property in all UE4 lights types called shadow bias. The documentation on this says decreasing the value allows the object to feel more grounded with shadows. This is pretty vague so I will give a little more insight on that.

From my experiments, decreasing the shadow bias increases the number of shadows your geometry casts. If you notice in the digital human UE4 project the default shadow bias is set to a value of 0.025. This is extremely low and it allows for much more accurate cast shadows and aids so much to the realism of the figure. All default values for lights are normally set to 0.5. It should be noted that if this value is set too low you will give some shadow artifacts on your model if it isn’t that high in polycount.

For instance, on the legs of my model, you would start to see the edges of the polys that composed the leg because the shadowing was too accurate to the surface. The head, being higher in polycount, was fine at lower values so there was a trade-off that needed to happen. A value of 0.065 seemed to be the perfect compromise for me. This value played a huge impact on realism for the character as well.

I had two maps in UE4 that tested the shaders and textures on the character. I had a look development map, which was just a duplicate of the digital human map and then the Malta map which is where all the content for the video was. It was when I imported the model into the Malta map that I became aware of a disconnect in the lighting.

As I said earlier, the default value of the lights is 0.5 vs. the 0.025 lights in the digital human map. The body looked fine, but the detail on the face was really blown out when compared to how it looked in the other map.

 

I then set the area lights to a shadow bias of 0.025 and the body looked really off and all the edges of the polygons could be seen.

I spent an entire day thinking that this was a smoothing groups issue. I reached out to Adam Skutt and he said it looked more like a shadow bias problem. Thankfully, this was the issue and not smoothing groups.

The other lesson I learned came exclusively from CGMA. Most of the time in the class you are just sculpting the portrait. This caused concern for me because I didn’t know if I or anyone else in the class was going to finish our characters completely. The class is called “Next Gen Character Creation”, so the name implies that you will be making a whole character in 10 weeks time. As it turns out only two people have done this in Adam’s class during his entire time of teaching. This put to rest any nerves I had about not being fast enough.

I was thankful we got a lot of time to finalize that sculpt and apparently, this was the norm in the class. Some portrait sculpts apparently can take up to 6 months in the industry according to a few people I have asked, so it didn’t seem unheard of to spend the majority of a 10-week class sculpting the head. Most of the mind-numbing sculpts and works I’m seeing nowadays have taken around 4 months to complete. These numbers are coming from the heads of the industry too, so even people of such a high skill level require the proper time to make amazing works of art.

With that, I conclude my breakdown for my character, Amasol. I hope I was able to answer all of your questions thoroughly enough to where there aren’t a ton of question marks.

  

 

Real-Time Character in 3D

Interview with Gue Yang

Gue Yang took CGMA’s Next Gen Character course led by Adam Skutt and talked about creating the first real-time 3D character: Mari & ZBrush workflows, solving issues with export to Marvelous Designer, a few useful tutorials and more!

Real-Time Character in 3D

 

 

 

Introduction

Hi, my name is Zhuyaj and I’m originally from the Bay Area, California. Currently, I’m a character concept artist at Turtle Rock Studios. I went to school for Illustration and Fine arts and pursued concept art in the game industry in 2011. I’ve always wanted to make a game myself but thought that I was a poor designer/painter so I opted for art school to sharpened that aspect first. Although I’ve barely scratched the surface of the 2D realm, I feel that there is some light of confidence in jumping into the world of 3D and finally doing what I’ve always dreamed of.

 

Start of the Project

As a word of warning, when I decided to take CGMA’s Next Gen Character art class taught by Adam Skutt, I had no clue what I was doing as it was my first ever 3D character and learning every app from scratch was a brain killer! All I knew was that I wanted to do a fan art of the classic survival horror RPG released in 1998 by Squaresoft, Parasite Eve. However, it would have to be re-imagined in the current next-gen realistic style.

 

 

In the game Parasite Eve, you play as Aya Brea, a rookie cop in the NYPD who fights off a viral mutation that spreads across NYC. While I looked for a current real-life replacement for Aya, I researched current detective shows that portrayed a confident young female detective:

 

 

After gathering dozens of inspiration reference, I started the concept art. I made sure to take elements that I wanted such as the bomber jacket and jeans, and detective badge and a gun holster. Among the changes in the design was the badge necklace, plus I gave her an inner hoodie:

 

 

I made the concept mainly for mood and feel rather than likeness, and I ended up making the final character a bit older.

 

First Steps in ZBrush & Mari

In ZBrush, I started off with a low poly base mesh that would match the concept’s proportion and body type.

 

 

Next, Adam recommended that we do a ZBrush quick sculpt of how our clothes would look/fit on the body before going into Marvelous Designer:

 

 

As for the clothing parts, I did quite a bit of reference research to find the parts that I needed. It was mainly doing the proper research on the specific piece of pattern and then I started right away in Marvelous.

 

 

Exporting from Marvelous to ZBrush

The only issue I had was connected with exporting from Marvelous Designer. Due to the issues I had when merging and ZRemeshing, I used a method shared by character artist Yuri Alexander. He imports separate adjacent parts and thickens them by groups. Then he merges them all at the end. 

 

 

By week 3, I had imported the simulated clothing from Marvelous into ZBrush. The inner parts of the hoodie which aren’t seen were deleted since they were unnecessary.

 

Head Hunting

Because this is a 10-week course which requires learning new programs, Adam asked us to aim only at getting the bust done. However, I was spending about 8 hours a day after work on this character, and even then I finished only by the 11th week!

To get the likeness I gathered tons of reference of the actress I chose, Lauren German. I didn’t want a 100% likeness, so I decided to omit certain features like the lips, jaw etc.

I divided the head refs into frontals, three quarters, and profiles:

 

 

Base Sculpt

 

With pore details

 

 

In Mari, I used XYZ displacement pores projection. There’s a ton of tutorials on dividing 3 main pore displacements into RGB channels in Photoshop, then project painting them onto the model in Mari. Afterward, you just export 3 Mari displacements back into ZBrush as layers, that way you can find the perfect balance in blending.

 

Hair Creation

For the hair, I followed Adam Skutt’s tutorial on real-time hair which you can find here: Realtime Hair Tutorial. The process starts off with generating Xgen hair in Maya for texture. As for the polyhairs, they consist of 4 layers of polyhairs in ZBrush. This was probably the most tedious task in terms of time. I took about 4 days hand placing each poly hair in ZBrush.

A quick Photoshop concept sketch to imagine the hair direction:

 

 

 

As I said, the most tedious task was placing hairs by hand and I felt like an idiot when I found out there were plugins in Maya that could do it for you… But I suppose there’s some love in it when you do it by hand.

 

 

In terms of heaviness, I wanted the hair to look good in final render so I had two low and dense poly cards.

 

Texturing

For texturing, I used Substance Painter. Painter already has a decent library of materials which I can use for everyday mundane fabrics and hard-surface so I did not invent any fancy custom materials. I just used the premade materials and played with the variables to get what I wanted:

 

 

Basically, I used the method of making ID maps and then applying materials to each segment across all my garment parts. This jacket gif is a good example since it has several ID separations and Painter works really well with dropping in custom materials.

In terms of getting the final render to look a certain way (for example, the jacket), make sure to have metalness turned on plus anisotropic setting like you would for the hair. It allows you to mess with the reflections there and you can get some nice sheen.

 

Painting the Face

For the face, I started off with XYZ albedo map projection in Mari then took it into Substance for final hand-painting.

 

 

I painted the head specifically quite a bit because it’s a lot more complex than the garments, making sure that I get the main hierarchy of color relationships on their proper face parts and of course, some brush speckle for detail. I like to bake the normals and AO in ZBrush rather than doing a bake/export batch in Painter like with the garments. In the end, this gives me more control and subtlety of the face due to higher res textures like 16k.

 

PBR Materials

 

 

The skin shader material was a tough one to get right, especially as it was my first time using Marmoset and trying to figure how everything works. It took many iterations of scaling the figure size in Marmoset and tweaking the variables in the shader plus post-processing to get to a decent look and feel.

 

 

I’ve attached some images of the settings I used for marmoset materials:

 

 

Rendering & Lighting

For my final composite renders, I made a screen capture in Marmoset with the rendering settings cranked up to the maximum. As for lighting, I went with the classic single spotlight with two supporting rim lights behind. I wanted the final mood to be dark and horror/thriller-like so I chose a proper skylight that would support that.

 

 

Attached image shows some settings for Marmoset lights and skin shaders:

 

 

Feedback

I had a blast in the class creating this character and learning all the new programmes and pipelines. It is definitely a step up in terms of time-management and technical side when comparing it to 2D art. I love that I can basically teach myself anything if I look for the proper tutorial and ask the right people. The only issue is that everyone has their own way of doing and approaching every step so it just shows how vast the field is. I believe I lost about a week of work because I couldn’t solve certain technical issues, but things like that are a part of the pipeline heartaches.

 

 

 

To be honest, I like that PBR pipeline is getting more artist-friendly these days and I hope to see it advance even further. When I was researching what tutorials could help me in creating this character, I found Frank Tzeng’s likeness tutorial on Gumroad which helped me a lot with sculpting. Simply watching him sculpt was all I needed to be a bit more confident with the human forms. As a character art beginner, I am still learning and gathering knowledge so I don’t have enough resources to share.

Thanks for taking a look and I hope you won’t be harsh on it! It’s my first introduction to 3D after all. You can reach out to me regarding any questions as well as 2D art since that is my expertise.

Realistic Character: Accessories, Sculpt, Materials

Interview with James Chan

James Chan did a breakdown of his realistic character Biker Girl made during CGMA course Next Gen Character Creation led by Adam Skutt: hair, skin texture, materials, clothes, rendering, and more.

Realistic Character: Accessories, Sculpt, Materials

 

 
 
 

Introduction


I’m a character artist and I’ve been working in the VFX/animation industry for about 8 years now. I’m originally from New York City and grew up inspired by Looney Tunes, Disney Animated Films, and Pixar movies. Those things probably led me to pursue animation and film as a career.

After graduating in 2010, I spent a few years 3D modeling for TV commercials in New York. Then in 2014, I joined Sony Pictures Imageworks and got to work on movies like Spider-man: Homecoming, Hotel Transylvania 3, Ghostbusters, Kingsman: The Golden Circle, and The Angry Birds Movie. I recently started working at Mr. X Toronto as a Character/Creature Artist.

 

Biker Girl

 

    

 

 

 

CGMA Course: Goals

 

After 3 years at Sony Imageworks, I started to consider jumping into the games industry. I was browsing Adam Skutt’s ArtStation and found out about his CGMA course Next Gen Character Creation. I had been a fan of Adam’s work since seeing his characters in The Order: 1886. I love the skin detail he’s able to capture in his textures and materials. So if there was anyone I wanted to learn about game characters from, it was Adam.

I had several goals for the CGMA course:

  • Understand low-resolution game topology, which is a bit different from high-resolution subdivide-able geometry in a film.
  • Improve my skills in ZBrush, Marvelous Designer, Substance Painter, and Marmoset Toolbag. At Imageworks, modelers mostly use Maya and Mudbox so I was out of practice.
  • Improve my texturing skills since Game Artists are expected to model AND texture. I only modeled at Sony so this was a weakness I had to address.
  • Produce a portfolio piece that could help me secure work as a game character artist.

 

 

    

 

 

Choosing a Character


I noticed a lot of my characters were male and stylized so I went into the course wanting to do a realistic female. I recommend every artist try to tackle projects that will teach them something new. Adam gave us a lot of freedom to choose what character we made. When he asked us to find a concept for homework, I literally just searched in Google “cool sci-fi woman” or something along those lines and found Heo’s awesome design!

 

First Approach to a Realistic Female Model


Sculpting women is more difficult than sculpting men because your margin of error is smaller. However, this character’s face is based on my favorite actress, Judy Greer, so I was quite familiar with her face already and enjoyed sculpting it. For reference, I would gather a bunch of images together in Photoshop and arrange them by angle, from front to side. That way, I didn’t need a million different photos on my screen and I could flip them horizontally to get the opposite angle.

 

 

 

 

A lot of the realism is achieved by displacement maps from texturing.xyz. There are a lot of tutorials on applying those maps, but I recommend Salim Ijabli and Jonas Skoog‘s ones since they only need Mari and ZBrush. What’s important to remember is that Texturing.xyz alone is usually not enough to get the best results. I would recommend at least one surface noise pass in ZBrush before you apply the displacements. And if you want the utmost realism, you’ll still have to mix in some hand-sculpted wrinkles, blemishes, and pores.

 

 

 

 

Clothes & Accessories


The clothing is done mostly in Marvelous Designer. Skin-tight outfits don’t benefit too much from Marvelous Designer since there isn’t much to simulate, but the program was still useful for the subtle folds on the jacket and sleeves. The toughest part was retopologizing all the folds and wrinkles into a low-res mesh, which I did with Maya’s Quad Draw. Thankfully, it didn’t need to be all quads, since triangles are okay for real-time meshes.

As for all the accessories and details, I approached them as individual mini-projects. Each day I committed to finishing just one item: a boot, a glove, a belt, and so on. The real problem is trying to bake the normal maps since there are so many overlapping or intersecting parts. Marmoset Toolbag and Knald are great for this because they allow you to organize your bake into separate groups. So even though the pants and belt intersect, their normals can be baked onto the same map without any projection problems.

 

 

 

 

 

 

 

Hairstyle


I followed Adam Skutt’s technique of making hair cards, which can be found on his gumroad. The basic gist is that you make small planes in Maya and grow XGen Hair from those planes. Then you do an Arnold render from the front view to get the Hair Texture. You need different kinds of hairs to give you variation in length, thinness, or type (like eyelashes, eyebrows, stubble, etc.). As for the cards, I just manually placed them. Before this project, I had never used XGen, rendered with Arnold, or ever placed hair cards.

I was going for a light and soft “modern bob” hairstyle. That hairstyle involves a lot of hair strips that split up the outer surface and makes the hair feel light, which lends itself nicely to haircards! It really helps to include fly-away hairs or singular hairs that messily come off the head to add some realism. To finish it off, you have to surround your hair with a sphere and then project the normals onto your hair cards so they don’t look too much like cards (check the end of this tutorial). I’d also recommend giving your hair a bit of subsurface scattering in the render.

 

 

 

 

 

 

 

Materials


For the materials, I leaned heavily on Substance Painter’s default materials and adjusted them as needed. For instance, most of the leather is based on the ‘Leather Bag’ material and the metallic parts were “Cobalt Pure” or “Aluminum Brushed Worn”. Combining so many different materials onto a single texture sheet normally requires laboriously hand-painting masks to separate the leather, cloth, and metal. Thankfully, Substance Painter can bake ID maps based on polygroups in ZBrush. So just organize your materials into polygroups in ZBrush, export the decimated hi-res mesh, bake the ID map in Substance Painter, and isolate your materials. Very useful!

 

 

 

 

Skin Texture


The skin texture is entirely hand-painted. I followed Magdalena Dadela‘s tutorial about painting skin textures with Substance Painter (see the video below). It involves using Painter’s “Dirt 1” and “Spots” Brushes to paint redness, freckles, and moles.

 

 

I also recommend looking at texturing.xyz’s cross-polarized photos. The pictures on the site are a great reference for how the skin looks without highlights and shadows, which is perfect for albedo maps. Finally, I recommend making a cavity map in ZBrush to target the pores and give them a reddish hue.

 

 

 

 

Rendering


Rendering was done in Marmoset Toolbag, which was a new tool for me. Navigating all of Marmoset Toolbag’s options was probably the most difficult part of the entire course. In fact, I still don’t know what a lot of the settings do. I’ll call attention to the most important tips for lookdeving, lighting, and rendering a character in Marmoset Toolbag:

  1. Group your lighting setups so you get specific lighting for different camera angles and poses.
  2. Use many cameras and lock them into position when you find a good angle.
  3. Increase the width of your lights to get smoother shadows. Shadows will be jagged if you don’t.
  4. Set the Environment light in a Marmoset Toolbag scene to be very low. It looks really fake so rely on your own lighting setup.
  5. When lighting, don’t go too crazy with getting specular highlights on everything. Sometimes less is more.

 

 

 

 

Feedback


I believe the most difficult challenges during this project were the ones I placed on myself. I happened to be wrapping up two other characters at the same time I was working on this character for Adam’s course. As a result, I fell behind during the class and didn’t actually finish this character until about 3 months after the course had ended!

Before this, I always felt I could learn everything by myself through video tutorials. I was skeptical of workshops like CGMA, but this course changed my mind. Getting to see a pro like Adam go through his entire workflow, from beginning to end, showed me so much that I didn’t know. Also, the weekly schedule and your fellow students keep you motivated to reach the next milestone of your project. I would highly recommend this CGMA course to other character artists.

Turtle Clan Warrior: Real-Time Character Production

Interview with Harley Corke-Ogg

Harley Corke-Ogg talked about the production of a real-time character Turtle Clan Warrior within CGMA's course Next-Gen Character Creation mentored by Adam Skutt.

Introduction

Hi, my name is Harley, I'm a Character Artist currently living in Ireland. I have three years of experience in the industry and currently working as a freelance artist for Scott Eaton. Before that, I worked at a game studio here in Ireland as a Character Artist. 

Before that, I worked as a fabricator in Scotland. I was fortunate to work on some pretty cool projects from designing and building bespoke furniture for coffee shops to sculpting and fabricating for artists. Even though I was delighted to do that kind of stuff for a living, I was always a secret nerd at heart and realized I needed to pursue my passion.

Goals

I decided to take CGMA's Next-Gen Character Creation class after seeing Adam Skutt's artwork and his hair tutorial. I knew he was someone I would love to learn more from. 

My main goal was to up my skill set and create a piece that could fit into an AAA game. I also wanted to get a realistic real-time character portrait which is something I really admire in Adam Skutt's work and learn how to push the facial details under the constraints of a real-time render engine. Apart from those goals, I looked for an opportunity to focus on getting better at organic modeling as a whole.

Turtle Clan Warrior

 

 

 

 

 

 

Concept

I found Adrian Smith's concept art when I was looking through Kotaku one day. There, I saw an article about the board game Rising Sun and instantly fell in love with all the artwork and miniatures. After doing a bit of research, I found out that Adrain Smith created all the concept designs and illustrations. At that time, I had no idea that Adrian was behind many pieces of artwork for Games Workshop that had a significant influence on me when growing up! It's no wonder why I was drawn to the design of Rising Sun.

Another reason behind my choice is my background. I was born and grew up in Singapore and I have always loved Asian mythology and culture. Naturally, I clicked with Adrian's concept art.

Finally, I wanted to practice both creature and human anatomy. Taking this into consideration, I decided to move away from the concept slightly and find an actor who I could use as a general reference to help add more realism and detail into the model. In the end, I decided to use Jeremy Holm as my primary human reference; I felt he had many subtle similarities to the concept like the underbite, similar nose shape, and the facial expression.

 

 

Modeling

My modeling workflow was pretty straightforward. I started by roughing out the design sculpt in ZBrush using DynaMesh; once I was happy with the proportions and silhouette, I started cleaning the model up. I ZRemeshed it so that I could use subdivisions, smoothing groups, and layers. Finally, when I was happy with the model, I exported a decimated version into Maya for retopo and set up my naming conventions so that I could export an FBX and use Match by Name function in Substance Painter to get a nice bake.

 

 

Skin Texturing

For the skin, I used texturing XYZ maps for the displacement and as a base for the diffuse. I projected these maps onto the model in Mari and then imported them into ZBrush. I applied the displacement on a layer and used a very weak morph brush to integrate the details into the face sculpt. I masked the details by the cavity and then inflated the model very slightly, which helped to incorporate the scans a little bit more. I then made a separate layer where I added in all the secondary wrinkles and imperfections.

Clothes

I used Marvelous Designer to do a quick simulation for the trousers, then in ZBrush, I used  DynaMesh to chop the mesh up so that I could rearrange it and resculpt areas I didn't like. As I enjoy practicing sculpting cloth, for other areas like the belt and the fabric on the feet, I quickly took some pictures on my phone for reference and sculpted those pieces by hand.

 

 

Rope & Beads

I found the rope pretty hard to make as I didn't know the best way to go about it. Eventually, I settled on creating a tiling rope material that I made in Substance Designer, which I could then bake into an IMM brush and use in ZBrush. This way, I could quickly rough out the rope placement using Z-Spheres and use the low poly loops to create the curves that would place the rope geo. Then all I needed to do was apply my tiling material, and it was good to go!

As for the bead patterns, years ago, I bought a book that had lots of Chinese textile and ceramic designs. I selected ones that I liked and scanned them into Photoshop. Then, I created masks that could be applied in ZBrush using the UVs. In ZBrush, I masked and inflated the model and used morph targets and layers to sculpt out the seam lines.

 

 

Hair

I created the hair using XGen and Maya following Adam Skutt's hair tutorial from his Gumroad. After I finished the hair texture in XGen, I started off creating rough proxy hair in ZBrush, which I could import into Maya. I then created the hair cards and a series of dreadlocks out of those cards using deformers. I decided to hand place all the hair cards and dreadlocks in a set of layers as most of the artists I researched seemed to all agree that it gave the best results over any plugin.

 

 

Secondary Details

All the secondary details were either added in by hand or added using NoiseMaker in ZBrush. I stored the surface noise on a layer that I could then further edit using the morph brush. This technique helped me to apply loads of detail and break up pretty quickly. However, I decided to leave out any micro-detail as I could add it in later in Substance Painter or Marmoset.

 

 

Materials

I love using Substance Painter and enjoy building my materials from the ground up. I find it helpful to work on each channel individually in the viewport and gradually add more breakup and depth using different masks like smart masks, AO masks, cavity masks, and generators.  However, I left most of the small micro-details out as I planned to use John O. Owsment's custom tiling Marmoset shaders. This material allowed me to use color id maps to create masks for micro details that I added in using tiling normal cavity maps in Marmoset.

 

 

Presentation

I chose Marmoset for rendering as this was the program that Adam Skutt was using in the workshop. I wanted to take full advantage of all of his teachings; so it was a pretty easy decision to make. Up until that point, I had only used Unity for work, so I was more than happy to try a new real-time engine and found Marmoset to be very artist-friendly and easy to learn.

As lighting isn't my main focus, I tried my best to keep it simple and show the model clearly with a simple three-point lighting setup with multiple rim lights to help the silhouette pop.

Feedback

My biggest challenge was to stay motivated and focused. As I wanted to push past the quality of my previous work, I concentrated on iterating changes through the pipeline as many times as needed until I was happy with the result, which was rewarding but also pretty time-consuming.

Overall, I really enjoyed the course. The tutorials and feedback were super beneficial, Adam is a great instructor. From what I saw, he gave excellent feedback and encouragement to all the students no matter how much experience we had. He also pointed us to loads of great resources like Peter Zoppi's eye tutorial, his own hair tutorial, and John O. Owsment's SkinDetailShader. This meant we could spend more time going over other areas of his workflow. But above all, it was super motivating and inspiring seeing all the excellent work all the other students were doing!

Harley Corke-Ogg, Character Artist

Interview conducted by Ellie Harisova

Realistic Character Art: Face, Clothes, Tattoos

Interview with Juras Rodionovas

Juras Rodionovas talked in detail about the production of his ultra-realistic character Soulless made during CGMA's class Next Gen Character Creation with Adam Skutt.

Realistic Character Art: Face, Clothes, Tattoos

Introduction

Hello! My name is Juras Rodionovas and I currently work as a Character Artist at Fatshark, in Stockholm, Sweden. Previously, I worked at Avalanche Studios where I got to be part of the projects Rage 2 and Generation Zero. 

My interest in 3D art started about 10 years ago when I was a kid, and got into modding games such as Fallout 3 and Amnesia: The Dark Descent. I suppose you could say that it was an introduction to game development and the varying type of work that existed within it.

I tried out a lot of different things on a surface level back then - scripting, level art, animation, and 3D. It gave me an overview of what happens behind the curtains of the game development process, and eventually, it led me to want to pursue it as a career. 

I knew that I wanted to focus more on the creative side of the field, but I didn’t know at the time that it would lead to character art specifically.

 

 

          

 

Taking a Course at CGMA

I was always interested in taking a course at CGMA to improve my skills since I knew the courses were being taught by world-class industry professionals. There were quite a few courses to choose from within character art, but I always have been a big fan of Adam Skutt’s work and the quality of it. So I decided to go with Next Gen Character Creation Mentorship and try to get most of it. By learning Adam’s thought process, workflows and getting feedback throughout the course to push my abilities further.

My main goals with the course were to update myself with the latest workflows and pipelines when it comes to character art for games, and also push my quality bar across various areas within the craft. Anatomy, realism, texturing, getting shaders to look good, and presentation. I also wanted to get an insight into what is coming up on the horizon of real-time character art, since next-generation games are just around the corner. 

 

 

 

Marmoset scene with all models, textures, and shaders assembled

 

Soulless: Inspiration

This character was actually based on a Korean fashion model and tattoo artist Marely Hong (also know as Kim Wooyoung), who passed away a few years back. I found a portrait photograph of him when I was browsing Pinterest one day while I was looking for inspiration/references. His look really stood out to me, and I simply loved all aspects of this person and his style. He had a really interesting and intimidating facial structure, a really cool taste for fashion and contrast in fabrics. Using jewelry and piercings to accent it. Lastly, he had really unique looking tattoos which I thought would be a lot of fun to try and recreate on the model.

 

 

 

 

Main references of the Korean model Marley Hong that I used throughout the project.

 

Another reason why I went with this person as my main reference for the project was that I really wanted to add some diversity to my portfolio. Creating a character of the non-caucasian race and having a look with so many tattoos that is very uncommon to see in daily life outside was a very attractive idea to me, and I saw it as an opportunity to challenge myself too.

I personally believe that there is quite a lot of generic-looking characters in the world of game art and video games. Seeing someone be able to create a diverse set of characters and faces is something I really admire, especially when they can portray some kind of feeling, and add history to the anatomy. To me, that shows a high understanding of both anatomical and artistic skill, and it makes the characters a lot more interesting to look at. It is something that I believe requires a lot of studying, and observation of life and people.   

Initial Modeling

When it comes to modeling and the production of this character, I used ZBrushMarvelous DesignerWrap, and Maya. For the texturing, I relied on Substance Painter

I started out by using a base mesh in ZBrush to sculpt on, which gave me decent topology to start with. The base mesh had a body too, which was important for me. It provided the ability to match my reference’s body type and get the clothing to sit the right way in Marvelous Designer. 

 

Initial body mesh that was based on the “Nickz_humanMale” base mesh that can be found in ZBrush

Modeling the clothing was quite a bit of a challenge, and it took me a few tries to get the jacket just right. Marvelous Designer is obviously huge help, but you still have to do research on how the patterns look and are sewed together to make it look believable. Getting the biggest folds and shapes is the hardest part in my opinion, and the detailing process is usually very straight forward. That is why I usually spend most of my time on the first stages of character modeling. Even if the detailing work is really good, it won’t save the model if you have a bad foundation. The sooner you can realize that the sooner you will be able to plan your work, and have a clear oversight in your head ahead of time. 

Geometry

To get my final geometry on the head, I used a base mesh that the class received from Adam Skutt and adapted it to my sculpt using Wrap3D. It is a very intuitive software that makes it really easy to re-use geometry on heads and even bodies. Once the new geometry has been applied, I could go back to ZBrush and re-project all the work that I had done on the starting geometry. This meant that I could start working on finalizing the forms on the face and start with the detailing work.

 

New low poly geo with UVs was transferred on the initial sculpt using Wrap

As for clothing, since this project wasn’t made for game production, I used the quaded geometry that you can get from Marvelous Designer, and simply used my 2D patterns from the software as UVs. This also made sure that the tiling of the fabric details behaved as expected. 

 

Re-using and setting up sewing patterns as UVs in Marvelous Designer before exporting the quaded mesh

Face

Sculpting faces is one of the most fun parts of the character for me and is usually an iterative process in my workflow. In this case, my character had some really subtle overlapping forms, that I felt were the most tricky to get right. My reference having a lot of face tattoos also made it more challenging. I had to really rely on my anatomy knowledge to fill in the gaps. It’s always a bigger challenge when you’re working on a likeness. But I feel that the reward of completing a project like that is very rewarding, and it expands your visual library. Which becomes useful when creating faces in the future based on concepts or from imagination. 

 

Having anatomical knowledge is helpful in situations where it can be hard to read the form in reference; it helps locate crucial landmarks, read the skull structure and the relationships between the planar structures of the face

I like to begin by working with relatively low subdivision, and solely focus on the primary shapes and proportions first, without getting distracted by details or smaller forms. This a very important step because it lays the foundation for the upcoming stages. However, since I always preserve my subdivision history, I can always come back and make “bigger stroke” changes, even if the detail has been laid down.

Getting all the big shapes and proportions down on the first try can be really hard, so I like to take a small break after this initial process, and come back with a fresh perspective or after getting some feedback. It always helps me see areas that can be improved, and push the face further. Almost everything I do on the head is an iterative process. This means that I don’t need to feel as if I have to nail the sculpt one hundred percent on the first try, and usually end up going back a lot once I have initial textures and rendering scene setup. The faster you can get to the final look of your presentation of the model (with shaders, textures, lighting, etc.), the easier it is to iterate and nail the look you are going for. That is because it gives a wholly different context and perspective compared to viewing a gray model in ZBrush.

 

 

     

 

I believe a lot of people think that it has to be this very locked down process and that each step needs to be final before moving on to the next one, but when it comes to a production environment and reaching the best possible quality, iteration is key.

Tattoos

Tattoos on 3D characters tend to often look very fake and “projected”. At first glance, it can seem that a tattoo is a layer on top of the skin, and only has a black color (unless it’s colored). But in reality, tattoos are scars, and the pigment is under the skin. This means that once a tattoo is healed, you no longer see the outer layer of the pigment within the skin, and the underlying ink that is visible, is even deeper under the skin. It becomes more blurred out because of this, so when texturing tattoos, it’s important to remember that it shouldn’t just be a layer on top of your skin layer with a clean and sharp projection.

 

A close-up picture of a face tattoo; you can see how blurry the ink underneath the skin is

You want to reduce the opacity and blur out the tattoos quite a bit. Hand painting helps because you naturally produce variation in the opacity and sharpness of the tattoo with your own brush strokes, leaving areas where the skin is more visible and where the ink is deeper. Tattoo artists are not machines, and when they are tattooing someone, imperfections are bound to be made, so you want to reflect that in your texturing work as well.

 

 

 

When it comes to color, you also want to make sure not to saturate the hues too much, because again, the ink is under the skin, and eventually the color of the skin tone will overlap the colors of the tattoo and make the saturation feel more muted and faded.

Clothes

Before I begin any clothing work, I like to do some preparation in the form of collecting references and researching to get the best understanding of the materials and how the garment is structured. This step helps me be able to understand what kind of 2D patterns I will have to create in Marvelous Designer later on, and how those patterns will be sewed together. This is probably the hardest part when it comes to learning how to create clothing inside Marvelous since the actual program itself is quite simple and easy to understand. 

 

Gathering relevant references and research in preparation for cloth production helps solve a lot of questions in advance

Sometimes it can be hard to find references of how the 2D patterns should look like, but once you have made a few garments, you start to build up your own visual library and can somewhat guess how the patterns should be constructed just by looking at garments and finding stitching lines on the references. A great way to get into Marvelous Designer is by first learning how to make basic clothing such as shirts, simple jackets, pants, jeans, etc. Once you memorize the patterns of that type of garments, you can pretty much make anything else with a lot more confidence since those basic garments always work great as a starting point once you have to make a more complex garment. 

Once the initial patterns of the garment are laid down, I like to make sure that the silhouette and the major folds look the way I want to. This process usually involves me “tailoring” the initial patterns to match the body type better, and so that I get more/less volume in areas where I need to. For me, it is also important that I stay at the initial particle distance of 20 and don’t start to add details and fancy stitching lines during this stage. This is because eventually, the simulation inside the software can start to get heavy (especially when you start adding more and more patterns, and increasing the density of the geometry), and when it does, it can be frustrating to make any bigger changes to the garment.  

 

Roughly how far I took the jacket before starting to add more detail and increasing the geometry

My advice is to stay at low geo density as long as possible, and only start to gradually increase the density once you are 100% confident that the garment is sitting well on the character, and that the big folds are the way you want them to be. Only then would I start to add details and I only increase the geometry right before I finalize the clothing, and export it to ZBrush for a sculpting pass. Another tip is if you have a lot of different layers of clothing, split them up into different files right before you start increasing the amount of geometry, and doing the final simulation. As long as you have the big shapes and forms overlapping on the different layers of garments, you don’t need to simulate everything together with a very low particle distance. This helps increase performance and reduces the chance of getting nasty simulation artifacts that can be a pain to fix. 

Patience and no rushing is key when it comes to Marvelous Designer. I had to learn it the hard way. 

Presentation

For the presentation of the character, I chose Marmoset Toolbag, mainly because it gave me the ability to focus on achieving as realistic presentation as possible, without having to worry about the technicalities of setting up a good rendering scene and shaders. I do enjoy working in game engines like Unreal Engine 4 sometimes, but I do that mainly when I’m trying to improve or learn something more technical.

When it came to choosing what kind of lighting I wanted to have for this particular character, I wanted to present it in multiple scenarios that could showcase the realism of the character in various ways, and also portray the kind of emotion and feeling I felt from looking at the reference - clear, bold, and confrontational. How do you portray that with lighting? Well, my thought process was to make the lighting natural, subtle use of rim lights, no unnatural colors, and not too much shadowing. 

The main setup presents it in a clear and neutral light, to help show the model as if it was photographed in a studio. This setup was also the one where I spent most of my time adjusting the lights to get the feeling I wanted on the character. 

 

For this project, I relied only on directional lights; in my main lighting setup, I only used three  - one key-light, one backlight, and one side rim-light

The other lighting scenarios were meant to present the character in a more natural way to prove that the textures and shaders work well in any environment scenario. The colors of the lights were sampled directly from the HDRI images, and the only thing that I adjusted on those lights was the size (in order to get soft shadowing). 

 

 

 

Challenges

The biggest challenge during this project was definitely the accuracy and patience required in getting all the elements of the character just right. Having an iterative process and not being afraid to take a step back in order to move two steps forward definitely helped. This requires quite a lot of patience but asking and receiving feedback from my mentors helped me get a new perspective every time I felt that there was nothing more I could push further. This is crucial because these last small steps and adjustments that get squeezed out of you is what in the end makes you wrap a project in a good way, and also level up your skills. 

At some point, you become comfortable in being able to reach a certain level of quality and consistency in your work. What makes you push your work further is either surrounding yourself with great, more experienced people with different perspectives or having a really good mentor. That is why I decided to be part of this mentorship, and it’s something I will keep doing in the future to keep pushing my work. 

Juras Rodionovas, Character Artist at Fatshark

Interview conducted by Ellie Harisova