Character Production for Film: Workflow Overview

Valentin Erbuke talked about the way he approached his character project Jungle Hunter: sculpting, texturing, working on the face and the environment. Read his previous article covering texturing tips here.

Life Updates

Hello! Recently I have been busy trying to rebalance my life and allow myself a little break time every day to enjoy life outside of work! The most interesting thing I have done was joining The Mill some time ago. I really like it there. I also had the chance to meet one of the people that influenced me to join this industry many many years ago, Alessandro Baldasseroni. And now I get to work with him every day!

Jungle Hunter

The Jungle Hunter project was quite challenging. I had been wanting to do this great concept from Russel Dongjun Lu for a while but I believed I wasn’t good enough to do the concept justice. It is such an intricate character and environment! When I finally decided to do it, I split the character and the environment.

Approach to Sculpting

I wanted to approach the character as I would in production. I focused on the torso and started creating this elf looking character as if I was approaching a regular elf in ZBrush. I completely disregarded the wood look as long as I wasn’t really far in the sculpt. And even then, when I applied the wood treatment to the face or other parts of the body, I made sure to work on a duplicate object that I used as a coat layer on top of the main sculpt. This way I would be able to go back if I needed to. My fear was that I would lose the appeal of the character if I sculpted too much sculpt details. I wanted to keep my primary and secondary shapes to read and light well. I approached many objects in the same way. I worked on extra layers of wood and moss this way. Later on, when I retopologized everything I merged everything together. After reprojecting all the details on my final mesh, I work on blending the detail with the skin.

There is a lot of detail that was manually sculpted. I relied on photographs mostly for surface breakup information, but all the shapes were manually sculpted and then displaced.

I also tweaked a couple of brushes in ZBrush to get strokes that I liked. As I went and got more comfortable detailing shapes I also tried to find a process pattern that would get me to a consistent wood result every time. I did this because there were so many pieces that needed to look similar and I wanted to keep a consistent approach.

Overall the sculpt of the character went pretty smoothly. Because I sculpted more than what appears on the concept, I also had to design a couple of elements which was really fun. I tried to complete the design of the torso all the way to the waist.

Texturing

I also split the texturing of the character and the environment in two. In fact, I barely worked on the environment until I had a production-ready character. The first thing I did was creating a bunch of groups to isolate my shaders in Mari. I was comfortable approaching the leather or the metal, but I had never tackled wood before. It was a lot of research at first: I made it look too green, or too saturated, or too dark. I mostly procedurally applied some color corrected textures everywhere to get a quick look, then manually went and painted mostly everywhere to refine the first passes. Then I projected some more to bring some specific areas out. I paid a lot of attention to color temperature and brightness to complexify the texture as much as needed. I used the cavity map, the curvature map and even the displacement maps from Zbrush as masks in MARI to help me work on the shapes. I also texture on decimated meshes so that MARI can bake a great AO with all the shapes of the sculpt. Because I sculpted so much before texturing, it really helped and paid off in the end.

I quickly started setting up shaders in Maya and Arnold for my different material so I could test them out in a forest HDRI environment. It would help me figure out if the different color temperatures and the brightness were behaving as intended. It also helps a lot seeing if the detailing is reading well and if there is enough texture information. In short, it was really helpful to do render tests rather than just being in Mari all that time.

What I really enjoyed with the wood was bringing back some really small saturated detail in the texture that added a lot of life to the shader. I stayed really meticulous. Everything had to look really organic and realistic, which means an incredible amount of detail that has to make sense on the surface.  I also painted some masks in MARI that would help me lookdev right away, like the lips and the paint per exemple.

Working on the Face

For the face, in particular, I was trying to use high-frequency lichen texture to simulate pores. It was definitely something experimental because I didn’t want her to look human, but I also didn’t want her to look like she was made of bark only. Her hair would be wood, but her face and skin would be more organic. Having the glossy lips and some SSS really helped convey all of that.

The paint was really fun to do too. I’ve been wanting to approach paint for a while to study how it would sit on top of a surface. Having some yellow on the face also really helped to bring out the contrast of the entire character and put the attention on the face.

The eyes were also interesting. I was lookdeving them right away to try to get an interesting look and ended up doing them entirely procedurally inside of Maya.

I think overall, even with a lot of iterations, the character went really smoothly. I believe I fined tune my workflow with a couple new tricks that I learned at the studio, which really helped.

Working on the Environment

Once the character was done, the modeling and texturing of the environment were really quick. I then decided to start scattering Megascans and using IvyGenerator to really add complexity and life to the scene.

After I was done putting in the main tree in the scene, I felt like the scene looked really dry and dead. I started playing with lighting and fog, but I couldn’t bring my trunks to life. That’s when I turned to Megascans for help. I started scattering an incredible number of small plants, moss, debris, leaves and more so on the surfaces with MASH. This was actually my first time using MASH, my previous scattering were all made in Houdini.

I really like MASH, it does a lot more than just scattering and the fact that I could easily alternate between random scattering and hand painted positionning, while still being able to go in and fix each instance manually was incredibly powerful. I started doing a couple of scattering tests on a new scene and when I got the hang of it started working on the real environment.

The Megascans I used were mostly atlases. I would bring an atlas on a plane and start cutting out the geo that I needed into the plane while looking at the texture, then lookdev all of those new patches that I created, and use MASH to scatter them.

Megascans Data is really incredible and layering them on top of each other really brought everything to life. I think I ended up scattering something like 50 thousand small moss patches alone across the entire environment.

At the same time, I would try out IvyGenerator for the first time. I brought my main tree trunks in IvyGenerator and I used it to create some smaller branches and a lot of cards that I brought into Maya. Then I redid the textured and lookdev of the IvyGenerator geometry, to make it fit in my scene. On top of that, I used that new geo to generate more instances with MASH and Megascans. I am really satifsfied with combining those tools together to get the chaotic look that I wanted for my forest. I ended up duplicating layers of my forest so that I would create the depth that I needed for the fog to look realistic. At the end the scene was gigantic.

I had so much fun playing with MASH that I ended up using curves to animate flying leaves going through the scene.

Animation

For the breakdown, I scripted some tools for that project that helped me animate the movement of the forest putting itself in place. I always liked the way ILM do their breakdown and wanted to do something similar. I also had to replace all the shaders with gray shaders but I wanted to keep the transparency of the cards so I scripted some automatic shader replacement for that too. Scripting was fast but sped me up so much.

Conclusion

Overall, the character took me 2 weeks to do, from scratch to the renders, then I spend 3 or 4 days modeling and texturing the environment, 2 or 3 days scattering everything. The shot took me a solid intense three weeks of work to complete, and then I spent an extra week riggign the character, animating the leaves, the background had some small swaying animation to it too that I scripted, and the doing the breakdown. So total it would be 4 solid weeks of work.  What was challenging but mostly motivating was the amount of different thinsg I had to tackle on this piece. It was really motivating because I had to use quite a lot of different tools to achieve the best result that I could get. I felt really creative in my workflow and approach of the work.

Thank you so much for reading! I am glad to share more about this artwork. I also wanted to share that with Yiihuu.com we worked hard to put online a video tutorial where I go over all the steps of this piece and demo my workflow.  You can find it here.

Valentin Erbuke, 3D Artist

Interview conducted by Kirill Tokarev