Overgrown: Vegetation, Modeling and Texturing Workflow

Introduction and Career

I’m Trym from Norway. I live in Toronto with my partner and our two cats. Last year, I worked at SpinVR, doing modeling, texturing, and animation for VR and AR. A lot of the stuff I did is under NDA, but I did work on a VR game that I was OKed to show, plus worked a bit on BrioVR, which is an online 3D viewer where you can do VR and AR.

The last thing I did there was the first pass on the boss and animation as he is introduced. My former colleagues picked up where I left off, having retopologized and textured him.

That’s my only job in the industry so far, but I’m really glad I managed to flex my creative muscles in a variety of disciplines there. As for right now, I just got a job offer at a game studio in Toronto, so I'll be busy with that going forward.

About Environment Art

The environment art is really cool. When I play a game, I spend a lot of time being captivated by the environment I’m running around in. The storytelling in environments is a big reason I enjoy games as much as I do. But honestly, it just boils down to being really, really cool. Characters and creatures are also very impressive, however, there is something compelling about the vast amount of variation and directions to go in under the environment umbrella for me.

About Overgrown

Aside from two rough blockouts before deciding on this one, Overgrown is my first attempt at a complete environment (the VR game stuff came after/midway through). I liked the other two environment blockouts, but the scope was simply too big for me at the time, as I was inexperienced and had a lot to learn. Choosing a more compact scene seemed like the smarter move.

‘Overgrown’ is the end product of mentorship at Think Tank Training Centre in Vancouver. My mentor is Kevin Meek, an environment artist at BioWare Edmonton, so we had our sessions online. My first task was to give him 3-5 concepts to start from, write some thoughts on why each would be good, why each would be bad, and he would give his thoughts and we’d settle on one.

The Inspiration Behind the Project

As I mentioned earlier, the reason for settling on Wadim Kashin’s concept as a starting point was because it seemed like a more contained scene, but still had a variety of assets that would challenge me. I would also have to work on truly making the house the shining star. I also wanted to incorporate a very contrasting storytelling element to the scene.

I love sci-fi, I love alien stuff, I like horror (in theory but I am too scared to play horror games), so I had some ideas. 

After I had the house modeled (more on that later), I imported it to ZBrush and used a cylinder to live boolean a hole. Using screenshots from different angles, I did a paint over.

Wolfenstein and Dishonored are games that have some quite impressive destroyed areas, and Destiny: Rise of Iron introduced some really cool looking red mutant growth stuff. Taking inspiration from this, I sketched out an idea of how it could look. The problem with this alien impact and the level of destruction was two things: A huge object just crashes through your house from far above at high speeds. The destruction would be much, much larger, especially on a house built around the 1920s. And secondly, it revealed a lot. Attic, second floor, first floor, basement. That’s a lot of open areas that I would have to figure out, model, set dress, texture. I was trying to keep my scope somewhat small, and this amount of destruction unveiled a lot more than would be necessary.

So the new idea would be to have it land off to the side from the house (much more so than in this early paintover), which would let me have more control over the crater size as well as have some interesting destruction on the house.

My end goal was to make a house that could live up to the quality of current games and to have a surprising storytelling twist as you progress through the contained area.

Modeling the House

The modeling of the house itself wasn’t too hard. I looked at the concept from Wadim Kashin, analyzed the type of building, and started looking for similar types of real-world references. I landed on a Folk Victorian style from the Midwestern US. From there, I found some blueprints of similar style Victorian houses to get an idea of the whole picture. Unfortunately, when you search for old houses, they usually don’t show you what it looks like on the back, so I definitely had to utilize some creative freedom and make my own blueprint based on what I could see from the concept, from my real-world references and from blueprints from the same time periods. I didn’t need to have a functioning interior, because I never had any intention of making the interior, but I needed to make sure the room sizes and dimensions made sense.

Seeing as my wood siding would be a material I created in Designer, the walls and construction are very boxy. So a lot of the detailing comes from the brackets, trims, and gables, which are decorative elements. Douglas McLean, who sat next to me at Think Tank (and gave me a steady supply of cookies) had a really handy book about architecture from exactly the period I was trying to hit, so I spent a lot of time looking through and cross-referencing that I wasn’t straying too far.

Here’s an early draft with a paintover, trying to figure out how to make this design a bit more interesting. There are several design issues here. The tower sticks out as it does on the concept. I wasn’t a fan of this, as I didn’t see any real-world references that had a similar design. But I did like the asymmetry of his design, so I wanted to merge his asymmetry of the L shape with the simpler construction of the white Folk Victorian house I found.

I imported the different sections of the house as larger chunks as I wanted to have an easier time with painting my mask for decay and weathering while keeping doors and windows and railing as separate, smaller modular elements, as well as anything that could be easily placed outside or through the house geometry.

The complexity of the house is more in the destroyed area, as I wanted to make sure the cross-sections of the house were realistic, so I would model the framework, joists, brick nogging, plaster layers, floor planks… a lot of layers! Initially, I did it so detailed because I intended to do destruction simulation, but during this mentorship, I was already learning UE4, Substance Designer, modeling, sculpting, lighting, and particles, so I tried for a little bit to set up decent destruction, but after a day of not getting anywhere near the results of what I wanted, I decided that it would save me time and frustration to simply model the destruction myself. I already had all the parts necessary. If I didn’t have so many other software to learn at the same time, I would have spent more time and effort in learning how to simulate the destruction well, but I realized I would sink too much time into it. 

The main area of destruction is a combination of flat walls with the tiling texture on it, and individual planks jutting out/jammed into tiling texture. I would bring the house into Zbrush and start sculpting the alien growth, pushing it into the house and finding interesting shapes for it, and readjust the house for it. The silhouette of the house needed some work, so I exaggerated it to fall out more but had to be restrained as to not reveal too much, as I did not want to spend more time modeling or detailing the inside. The final alien roots were made in Maya, using curves and positioning them along with the original sculpt, and then extruding a lot of different cylinders. Because of the complexity of some of the roots, I imported 3 fairly large root shapes, which I just positioned in the world so that it fits perfectly, and then I added to that with several smaller modular root models.

The other assets were modeled like any other asset, I would imagine. There’s not a lot of complex stuff aside from maybe the foliage. I would google a lot for “vintage” and then whatever I was trying to populate the scene with. Seems to have worked!

My ladder is a simple model. It’s just a lot of beveled boxes! But I wanted to use the base wood material that I had made for my siding in different areas. I put some colors that made sense for the rest of the scene, so I imagined someone would put down a paint bucket with some dripping paint on the top of the ladder while painting. 

The white, yellow and red colors are the same white, yellow and red I use on the house and garage. The green… we don’t talk about the green.

Working on the Vegetation

My grass, most plants, and the ivy were modeled by me. It’s not an overgrown scene without some ivy!

Getting the density and variation of grass and ivy to look good was a little bit trial and error, but since they were organized in Maya, I could move, add or remove leaves as it fit me. I sculpted 4 variations of the leaves (top right corner) and would make these larger bunches that I baked down into cards and placed on my house. Karen Stanley (Kazperstan) has a great tutorial for making some “basic sexy ass foliage”, in her own words. Now just repeat that for everything! I didn’t do it for all my plants, though... For many of my less seen or plants that took up less screen percentage (white daisies and lavenders), it’s such small shapes and vibrant colors that I didn’t need very sexy sculpts. And then there’s also some foliage where I would find a megascan of some weeds, and then make the foliage cards using that. They also have some ready-made stuff, which can definitely speed up your process.

I also made some spiderwebs. The first version I did was just way too noisy and had too many shades all over the place. The results weren’t what I wanted, and it masked out stuff badly. These were all based on spiderwebs I found online and masked out the stuff around, so that might be it, too. Either way, I took some of the key shapes from that and basically did a paintover, simplifying it. Then I treated it like foliage, so I basically made cards the same way I would for a plant, and then I imported it to UE4 and made stuff look old and spidery.

For trees, I dabbled in SpeedTree, which is just the coolest thing ever. It’s a bit daunting at first, as the interface is quite unlike most other things I had used, but after fidgeting around and moving every slider just to see what it would do, I figured some stuff out! The great thing is that once you make something you like, you save that file out, and then you can start randomizing your parameters based on your previous tree. So you get lots of variations that aren’t too wacky but not so similar that it’s the same tree. If the tree is very different (birch tree vs oak tree), it might be easier to start from scratch, and then make your variations by randomizing again. It’s really powerful, and I feel like I just scratched the surface of what it can do.

This was a huge learning experience, so I wanted to spend as much time as I could figuring out good workflows for myself in terms of prioritizing quality vs. prioritizing quantity, and a nice middle ground. I think it’s important to find out when and where to spend your time when you have a lot of things that need your attention.

Texturing

One of the things I’m most proud of is my wooden siding texture, which was made in Substance Designer. Before this one, I had made some test wood materials, and some other simpler materials to grasp the UI and workflow, but this was the true test. I’m still not a very technical guy, I like to do things very visually, so I was afraid that Designer wasn’t going to be quite the thing for me. Turns out, I love it. I don’t always fully understand the inner workings of the nodes I use, but I understand what I’m trying to make the nodes do, and the fact that you can see it, right there in front of you, is just awesome. I think my approach to Designer, in the beginning, was similar to my approach to SpeedTree - hit all the buttons, try all the nodes, get familiar with what does what and how I can manipulate that to my liking.

The siding is 1) A tiling base wood texture that looks solid. 2) The wood siding pattern, which is very basic. A linear gradient with a gradient map to clamp some values. Put that through a tile generator, and then stretch it a bit in the horizontal to make sure it doesn’t have any weird artifacts. Wear down that edge so it’s not all straight planks. They’re old and wood warps and it’s very hard to maintain a good house. The standalone tile generator uses the same amount of planks (the parameters are shared between the generators, so I can update both), and has a random luminance variation so I can use it as a mask for various things, such as offsetting the wood texture per plank. Don’t want the wood to be tiling from Plank A to B to C!

And 3) Chipped paint. I had to spend a lot of time figuring out how to lift the lip of the paint in such a way that it wasn’t too intense, and wasn’t too subtle. Lots of tweaking. Paint in itself isn’t too tough to make. You can essentially just put color on top of your wood and reduce the normal intensity. Still want some notion of the wood underneath, but if you apply some coats of paint, it’s going to be quite thick. After that, get some dirt in there. Put some nails in it. Make it seem like it’s rained a lil bit maybe, so there are some vertical streaks going on? Work on big details, then go to small details. I feel like Josh Lynch or Rogelio Olguin would say something like that.

My paint layer is just a mid-tone grey with some variation in it, to add some texture and “paint strokes” to it. So when I’m in UE4, I can make it whatever color I want by masking out the painted area from the exposed wood area.

My base wood is quite colorful. I just love adding some color variation, because stuff isn’t just 5 shades of brown and grey. I think it makes for more interesting material, at least this one specifically. I have at least two more wood variations that are a lot simpler, but this one is the one I learned the most.

The texturing of the house comes from the magical shader that Sean Disero made while we were working in tandem on our respective scenes. It uses the red channel to essentially blend between as many textures as you’d like to. The green and blue channels are unused on my material, but Sean used his blue channel to add wetness. The idea was to use the green channel to layer in more dirt or whatever he wanted as well. All of these could be painted in on the same texture mask. If you painted on a 256x256 mask, you get 65536 pixels. That’s a lot of information to paint on!

Setting Up

The scene in Unreal is a lot of trial and error, but I had a lot of waypoints to spots that I wanted to focus on. I was playing around with composition and camera locations early on, to figure out what I could skip/ignore and what I needed to make look really good.

You can set waypoints by clicking ctrl+1 to 0 on your keyboard. Then when you click 1 or 5 or 8 (without ctrl) you instantly get over there. Pretty handy!

Start simple, get a good idea of where you want stuff to be (white boxing it, if you will) and then start adding to it and replacing your placeholder models as you go!

The lighting is a skylight, a directional light, and some fill lights here and there to add a little extra oomph to spots I found needed some more love. I also spend a lot of time in Detail Lighting, to just look at the lights and shadows and not worry too much about colors.

Here I have a spotlight that doesn’t cast a shadow but adds a soft light to the left side of the building. It was getting quite dark there and I wasn’t quite getting the look I was going for with just light bakes, so I decided to add something to just add a little to the area. I did this in three or four other places too, just to add some better focus. No shadows needed.

Same here, as I wanted the trellis and area around it to catch the eye a bit more. I wanted to lead the viewer towards it, to be curious as to what is behind the house if you follow the path there.

Lightmass settings… these worked for me. I had the static lighting level scale even lower and the quality and smoothness at better values, but my computer takes quite a long time to even try to bake the lighting for that, so I ended up with this, which yields me good results. 100 indirect lighting bounces isn’t really necessary, but it also doesn’t seem to up the bake times much, so why not. Lots of great insight about light mass settings on the official UE4 youtube and Tom Looman goes through it a lot.

In terms of Color Grading, I could have spent a lot more time there. White balance, global, shadows, mid-tones, highlights… that kind of depends on what you’re going for. But LUTs are nice to have.

This is the final image without LUTs. I think it looks alright, but it’s a bit washed out. So I bring that screenshot into Photoshop, tweak the levels and hue/saturation until I’m pretty happy, and then I transfer that info to a LUT (Look Up Table).

Some Tips 

General tips… Hmm. If you’re just starting out and doing this all by yourself, then find a smaller scene or a very contained area. My scene is a pretty good size, but I think even smaller than that would be better. If you’re a veteran, I dunno man, go wild haha! I’m pretty fresh to this, so I don’t really know what kind of advice I can give, but me personally I look for cool shapes and silhouettes, cool negative space, cool storytelling.

I think you should check out Karen Stanley if you wanna make basic sexy ass foliage. She also has a great video series on Artstation Learning. Tom Looman if you want to learn more about Unreal Engine. Consume the Lighting Academy videos that Tilmann Milde has on youtube. DEFINITELY learn about texel density from Leonardo Iezzi. Timothy Dries has some great blog posts going through a lot of stuff too. Chun Chun Yang recently started her youtube channel, Artruism Digital, and while she might not be entirely game focused, her process of texturing is something to learn from.

Once again, I could not have done this without Kevin Meek, who is alright. Yeah, he’s OK. Also, Adam Bodden, who steered me on the right path to what environment art is, like a caring father figure only some years older than me. And then also Sean Disero, who I just wish would finish his demo reel at some point.

OK, thanks.

Trym Roedder, Environment Artist

Interview conducted by Arti Sergeev