Hi Elliott, This is a great breakdown and very generous in sharing your process and insights, you came a long way from the vending machine days!
Are you planning on releasing the UE4 project to the public? Or only builds? I'd love to play around with it in the editor if possible!
We’ve had a talk with Zrinko Kozlica from Guerrilla Games and discussed his most recent personal project. It’s an awesome 3d model, which was entirely textured in Quixel and modeled in Zbrush. Zrinko gives the full breakdown of the project and talks about some tips and hints that helped him to create this amazing character, heavily inspired by Dark Souls franchise. The model is based on the concept by Timofey Stepanov.
Hi! My name is Zrinko Kozlica and I am an Austrian 3D Artist currently living in Amsterdam, where I joined Guerrilla Games as Junior Character Artist to work on their new IP, Horizon Zero Dawn. This is the first game I am working on, so I was and am really excited to be part of such a big title. Before that I studied Information Design at the UAS Joanneum in the beautiful city of Graz, Austria and worked at Vienna based Motion Design studio EAT MY DEAR
After finishing other projects I was doing my daily browsing on Artstation as I stumbled across Timofey´s art. The complexity of the character and scene composition struck me immediately as well as the great artistic execution of it. As the last character that I worked on was part hard surface and part cloth I saw an opportunity to go into a more organic direction this time. The challenge of getting the different components right as well as nailing the overall mood were the factors that eventually hooked me. It would require sculpting a humanoid torso, somewhat believable cloth, placing cards for the feathers, creating the small diorama the character is set in and extensive texture work. When I started with the whole thing I realized the extent of the task, which was a bit overwhelming to be honest.
The sculpting and modeling part of the character needed some planning in order to have less backtracking to fix or change mistakes afterwards, even though the process itself was pretty straight forward. The sculpting was done in Zbrush and for some parts like the bells and gravestones I used Zmodeler to get the base shapes. For the cloth I used Marvelous Designer with a simple pattern setup for the basic definition and then brought it into Zbrush for further detailing.
The branches on his arms turned out to be surprisingly tricky. I tried different techniques to recreate the concept but no version could really satisfy me. At one point I wanted to create every small side branch with Zspheres like the main branches, but it seemed to be too much work, so I created the basic layout with Zspheres and used Dynamesh and the Snakehook brush to pull out the smaller parts. The result was far from perfect but still good enough to do the job, so I stuck with it. Looking back I would have done the extrusions properly by hand to have more control of it afterwards, despite the time effort, but I am sure that there must be a better way of doing this kind of things! For the feather cards i modeled a set of simple shapes in Maya and placed them procedurally in Cinema 4D, where I made a simulation setup for them to get some gravity into the flow. I used the same setup for placing the grass cards as well.
The UV layout and rig to pose the body were made in Maya using UVdeluxe and a bit of Headus UV Layout to pack the UVs. Here the option to group multiple shells into separate UV boxes before packing the whole set was really helpful to get some order into the layout of the different maps. For baking the AO and Normal maps a combination of the integrated Maya baker, Knald for fast AO and xNormal were used. I have to mention that after the Highpoly was finished the original plan was to do proper retopology of the whole thing to get a polycount of less than 100 000. At that point of the project though I already had invested so much time that I made the hard decision to use Decimation master and Zremesher for the most part, to achieve a polycount that’s usable in Marmoset and while texturing.
The biggest challenge with the materials was to manage coherency in coloring and material definition while capturing the mood of the concept art. Unlike shape and basic colors, the materials weren’t that defined in the reference, so there is a lot of room for interpretation and creative freedom but also for mistakes.
Again the original plan was to use a maximum of 4 materials for the whole character including diorama, consisting of 2k to 4k maps. That plan got scrapped when I switched from proper Game resolution remeshing and texturing to something in between. In the end it resulted in 10 materials in total, excluding the crow that is used as prop. A number that certainly could have been lower if I had put more time into the UV layout.
In the beginning of the project I had already decided to use it as test case for the Quixel suite, which I was really eager to try out. In Quixel I created a quick base color and material pass for all the parts. This helped me to get all the necessary maps and materials into Marmoset, so that I could check quickly how the changes I made worked together in an lit environment. Using preset materials as starting point helped me to iterate faster later on and to get a feel for the overall feel of the scene. During that phase the Quickmask mode turned into a good friend, since it allows to preview changes and the affected areas in a texture without the need to wait forever or risk a crash by previewing the final outcome in Photoshop,
After I was happy with the first pass of the textures I decided to add a projected texture pass and also some hand painting to emphasize the volumes of the skin and cloth mainly. The skin and cloth were definitely the parts where it took the most time to be satisfied. Something that really helped me not to get lost in insane amounts of detailing for closeups, was to focus on the size and distance to the camera in the final renders. Especially skin is really hard to get right for closeups, so I luckily could focus more on the overall definition then on the surface detailing.
During the late stages of texturing I realized that the character would need a lighting setup that really supported the overall mood of the scene in order to catch up with the concept. Luckily Marmoset is really easy to use, so the time spent in it is mainly to present something better rather than to figure out how to get a basic setup to work. The automatic reloading of meshes and textures is something that helps a lot to get satisfying results without long waiting times to manually reload content.
To start off I reduced the intensity of the Skylight a lot and did a basic 3 point setup, with a top left Spotlight, to get a calm composition for the eyes, a Rim light from the back to emphasize the silhouette and a fill light from the lower right side to counter the darkening shadows that were cast by the Spotlight. To my frustration this setup didn’t look the way I wanted and also made the character look worse then it could.
Fortunately a colleague pointed out that the concept reminded him of Bloodborne and Dark Souls. Having a closer look into screenshots of these games, in combination to gathering inspiration from other artwork, I got back into the right direction.
A more complex setup with multiple spots and fill lights, which got the specific coloring in Marmoset to get the warm, dreamy fantasy lighting, would light the character in a more interesting way. A big addition to the atmosphere was a feature just recently added to Marmoset. The fog element. This was the missing link between lighting and scene that, in combination with some tweaks in the camera post effects, finally made the whole piece work.
Working daily together with really talented people is a great opportunity to get feedback that really helps to make something better. Sometimes they point out mistakes and bad areas that you are already aware of, but hoped that nobody would notice, and sometimes they will see fundamental errors in the whole piece.
One of those fundamental mistakes was the noticeable difference of the sculpt to the concept art. Unfortunately it was already too late to fix too much in the basic shapes, since I wanted to get the whole piece done, so I stuck to fix smaller issues like cloth folds and the placing of some ropes in addition to having another pass of sculpting the hands, since they are really prominent in the scene.
In general I would say that getting proper feedback helped me to understand the character better and think more about why I make something in a certain way before I do it. There are a lot of ways to get a 3D model done,and many resources online that are accessible to get technical help. What makes personal feedback really valuable is that it helps to develop critical thinking and provides a different view on problems that you might or might not have seen. As long as you have this “Damn it, they are right, I knew it´s wrong” reaction to critique I would say you are getting good feedback and learning from it.
In conclusion I can say that no matter what happens to a started project, keep in mind that whatever the outcome may be: the experience you gather is something that stays and helps you to grow as an artist.I learned a lot on how NOT to tackle things with this character, but that was expected due to the complexity of the concept. Having clear goals helps in maintaining some hope when the temptation of just scrapping it and starting something new is present.
If the goal is to get a great sculpt then it’s obviously not necessary to think about retopology, UVs and materials. But if you want to go all the way with the sculpt then try think about making your life easier when you have the chance to. For example if you have things like the gravestones first make a low poly model, make a UV layout and then take it into detailing, otherwise you will have to redo the UVs and Bakes on the same thing multiple times and this can crush motivation. Also try to avoid shortcuts to things that just take time, in this case the branches on the arms. There was no way of getting a proper lowpoly mesh from the Dynamesh mess that it was without doing manual retopology, so I ended up with fairly high poly meshes from Zremesher instead of something really usable.
In any case just try to remember why you are spending your time on something. As long as you have fun and learn for the future it was time well spent, no matter what your original goal was.