Laurie Durand kindly talked about her experience of working in Ubisoft and taking part in the production of For Honor.
I’m from Montreal, Canada and as long as I remember I’ve always been interested in history and old architecture and loved to play video games. I did some private sculpture class for a couple of years and after that, I went to college where I specialized in visual art. After college, I built a portfolio and entered the university to study 3D art.
After trying many different things like animation, character modeling, and game design, I decided to specialize in environment art. I stopped my choice on it for good because environment art has more freedom for creativity and ownership as we always have to change drastically the basic concept idea to adapt it to the level design. And of course, you can be easily polyvalent in environment art (level art, modeling, texturing, shaders, lighting, etc.)
Later for my synthesis project, I teamed up with other 4 graduates and 3 programmers from other universities and we did a game prototype for the Ubisoft Game Lab Contest. The experience was a blast. We had so much fun doing our game and happened to win the price of the best art direction and 5 nominations in different categories.
First Experience with Ubisoft & Work on For Honor
I began to work at Ubisoft in spring 2016, so it’s been 2 years and a half since I joined the industry. In term of challenges, I was rushing so hard during the 3 years of my bachelor that when I joined the For Honor team, I expected to crunch again as the game was due to be shipped in less than a couple of months. In fact, I had the best integration ever without any stress and with a lot of time to finish my first task. The team was really caring and they wanted me to adapt well as I transitioned from being a student to a professional.
A thing I realized quickly was that the school taught us a lot of techniques and tips but there is a lot of specific stuff that school will never teach. For example, in For Honor, the camera can pass through the walls or buildings (camera occlusion) and when it does, you can observe the interior of assets rendered with a grey noise shader. For making this shader work properly, you need to close the edges of every single mesh otherwise it will show highlighted ugly holes when passing through them. The decals are the exception.
In terms of work, one of my challenges was to adapt to the AAA workflow in a team of a few hundred people. The communication quickly became a priority, especially in the environment art department where there were many artists of different specialties all dependent on each other. Then, a few months later after finishing For Honor, a brand new type of production task appeared: keep supporting the game that will go live. Every 3 months through the year after the launch we had to ship 2 new maps with different level design. That would mean way less time than usual for producing those maps and we clearly had to be more proactive and partly readjust our production pipeline after the 1st delivery. The first season delivery (Shadow and Might – Season 2) was at the same time my biggest challenge and one of my biggest pride since I joined the industry.
Insight into the Level Production
At first, I worked on 4 areas of this map. I had to put the right textures and set the right proportions for the proxy meshes. Then, once approved, 3 of those sections were sent to the outsourcing team while I kept the biggest and the most gameplay sensitive one: the green temple area. We wanted to make something less destroyed and more luxurious, so for this area, I did a more fanciful design than what we usually see in the game. Of course, that implied doing some iterations, using existing textures in a new way and modeling new ornamental parts.
Then, I received an inexisting map area that was really late in production: the tomb. That was an unusual and challenging task but I embraced it with pleasure.
It was the first time when I did not only polish modular assets and props but prepared all the basic level art layout that worked with gameplay from scratch. I also took a really big part in mesh placement and fixing the lighting.
After the first basic modular assets that made the area functional were ready, I started doing a rock-sculpted interior test. Unfortunately, because of the lack of good reference for Japanese rock crypts as well as the lack of time and resources for doing new textures or sculpts, the idea was aborted. Then we decided to go for plan B which was safer. I did a luxurious wood interior all carved with a few rock elements like the grave. Another new thing was to do a low-resolution vista not accessible to the player (a little shrine in the back of this room).
Once everything was finished and the map was validated by the art direction, I stayed on this map for fixing all the LOD bugs together with the level artist. We usually don’t have any problems with LODs as we use Simplygon plugin but in the case of this map, it was a concern because there were many large areas where you can see a lot of complex and organic shapes of the buildings at the other side of the map. So I had to rework topology of almost every asset in order not to lose important details.
Environment Design from Scratch
When I need to make something from scratch, in general, I begin with a brainstorm and I pick many pictures of visual references on Pinterest or Google. I usually look for pictures, historical references, videos, concept art for shapes, references for mood. I also look for some stuff that doesn’t directly relate to the basic subject, because it helps me to develop the very original idea.
After this step, I do sorting, choose the referencesscreenshots, assemble them in a moodboard and do some quick sketches based on the board. Depending on what I have to do, I do some sketch of composition or detailed props. It sometimes happens that I draw just to find a solution for an element which already exists in the level but isn’t artistically plausible enough or has some technical constraints.
At work, most of the time I skip all this step and go directly to sketching the environment design directly in the engine with no references or 1/2 references images already chosen by the art direction. Sometimes I draw just to find a solution for an element which already exists in the level but isn’t artistically plausible (for the world of For Honor, for example) enough or has some technical constraints due to the level design.
Once I get the approval, I readjust the scene according to the feedback of Art direction and clean the assets with the right metrics and gizmos. After that, I begin to better define modeling of the different elements and making high-resolution if needed in ZBrush or 3ds Max.
In the meantime (or after all the modeling is done), I do unwrapping. The next step is to do the first pass of materials, texturing and lighting. Since at Ubisoft I’m only responsible for modeling and level art, the textures, materials, and lighting are made by other colleagues. But if it’s a personal project, my process at this point begins to be more organic so I can go back to readjust the modeling or UVs depending on what I create. I polish the materials and textures, and once I’m happy with the result, I polish the lighting.
Importance of Grid
The grid is at the same time the first thing we determine in a project and the tool with which we work all the time. Determining the units of measurement in a project is important before starting steps like the construction of environment modules, textures, and level design. We must also take into account the environment’s proportions that will be credible from the player’s point of view and measures specific to the gameplay such as the minimum distance allowed for 2 players to fight freely. Once decided (usually by the gameplay team), it is important that all artists, animators, and level designers respect these measures. So the grid in both 3D programs and the engine is the most important tool for the environment artists to construct their modular assets.
The first thing to do for setting up the grid is to match it with the one in the game engine and its measures. For example, in Unreal 4 the units are in centimeters, so it’s recommended in the 3D software to set up the grid and the units the same in order to avoid scaling issues and ensure seamless exportation of the assets.
When you use the grid, the snap tool is a must for constructing in 3ds Max with the right measurements and the pivot (gizmo) at the right places. The location of the pivot point on the modules is important, because once in the engine the snap will allow you to place them quickly and on the grid just like Lego blocks. The grid is as well important when exporting because the pivot of the asset must be aligned with the origin of the grid (0, 0, 0) in order to avoid an asset with an offset pivot in the engine. Back to the engine, the snap tool will make it possible to build and iterate environments quickly with the grid.
Keeping the Balance
When I do the first pass of the basic modular assets I make sure that the gameplay flows well (by testing it) and get some feedback from the level designers. For example, a door entry in a third person fight game must be 2 meters wide so that the players don’t feel restricted when they enter a room or an area. Also, the open area for the fight must be at least 3 or 4 meters large. We also have to be careful with where we place the props or intricate details in the closed environments. The reason is that you don’t want to confuse the player if they need to react quickly. It’s also applicable to textures: our texturers try not to put too many colors or noisy details, and as a modeler, I have to be careful when I choose textures and do the unwrapping. Here, I also test the game to ensure that it’s playable and not confusing!
Approach to Assets: Sculpting Rocks
Doing assets true to reality can be difficult. For some elements like rocks in a lot of AAA projects, the team will envolve photogrammetry or will develop procedural technologies with displacement and/or tessellation in the game. But in the case of For Honor, a big part of the memory budget was already given to gameplay aspects like AI so procedural rocks weren’t the best option.
As we were a small but proactive team that got used to delivering new content regularly, we decided to work with a mix of sculpting in ZBrush and scans. The cliffs located in Jordan and in Monument Valley, US, were our main inspiration, so we searched for the cliffs pictures in those areas:
Once the chosen rocks were approved, we selected a 3D scan model that looked alike from our bank. Since the scans had a low-resolution it is easy to modify them. We cleaned the scan and roughly modified the general shape in 3ds Max. Sometimes we map the reference picture as a texture on it. The idea behind it is simply to get a topology with shapes and basic information about the cracks, not to obtain the best and most defined scan ever.
Then the biggest part of the work is the high-resolution sculpt of the scan. In ZBrush, I do a copy of the scan, close its holes and remove its exploded triangles. Then, I convert it into a high-rez dynamesh for keeping the definition as much as possible.
Afterward, I do a quick UV pass and project the scan’s textures and details on it. It’s not necessary to get all the micro details because it will serve just as a reference to start sculpting on. I refine the main cracks and other details of the rock by sculpting them, then move to the little details. I don’t need any additional references at this point and improvise with the remaining crack details. The idea, again, is not to reproduce the rocks 100%, but to get a rock credible at every angle and under any lighting.
Once the sculpting is finished I export the mesh to 3ds Max, do a low-resolution retopo with the PolyDraw tools, do the UV with the peel mode and bake normal maps. I prefer Marmoset for baking but you can use any other software solutions.
The main brush I use is the TrimSmoothBorderBrush with alpha from the online ZBrush library, a plain version of TrimSmoothBorderBrush (you can find it manually in AppPixologicZBrush 4R6ZBrushesTrim) and a Clay Build Up in spray mode.
Working on For Honor, I had the chance to make modular assets such as pillar walls and basement arches, plus some prefabricated kits of interchangeable fortress walls. For some of them, I only had to make sure they were functional before sending them away to be polished.
In terms of the workflow for the modular pieces, the first step would be adjusting everything to the defined grid as I explained earlier. The measurements of the modular piece should be exact taking the pixel ratio and textures into account. Usually, when the texture dimensions are multiples of 2, we will make modular assets that also have the measurements multiple of 2. For example, in a third person game, the pixel ratio could be 1 meter equal to a texture of 512, 2m = 1024, etc. and the modular pieces could be 2m, 4m, 8m long. Note that there is no specific rule for the pixel ration that a game should have. It’s different for each game.
Advice: always build an intact version first before going for more details (in a destroyed version, for example). Besides, keep in mind that it is very hard to make modules with organic and/or complex assets such as trees. As for modular rock elements, you can model them the way you want in order to achieve an interesting look at any side.
In a game like For Honor, there’s a lot of organic modeling such as destroyed walls, handmade and crooked Viking buildings and so on. Luckily, the laws of the gameplay constraints were less strict, so sometime we could make modular pieces 2,75 long, and sometimes because of the drawcalls limit we had no choice but to attach a bunch of modular pieces together.
Whether modular or not, all the meshes were constructed with modular details.
For example, at the beginning of the production, our lead modeler had to sculpt 2 kits of wood planks and construction stones, in cleaned and damaged version. The idea was to use these little details in order to quickly polish every modular architecture piece with a high level of details. Afterward, they were put in a common bank available to all the artists. These meshes were used for all assets in the game in any possible way whether to make a hole in a wall or a bridge. Of course, the more we advanced in the production, the more specific needs appeared. Therefore, we had to model other versions of broken planks or rocks and add them to the bank just to have more details available for polishing.
As you can see, modularity is suitable not only for level art but also for modeling the modular pieces themselves. It’s useful for the sake of time. The only issue with this workflow is that you need to organize and make a bank early in the production.
Personal Advice for Learners
For someone who aims to become an environment artist, I would recommend doing drawing or painting, sculpture, photography, art classes, and so on. In short: develop your artistic side and your artistic eye. Just watching new art on websites like ArtStation, going to an exhibition or even discover beautiful places (if you’re lucky to travel a bit) can help. All these things will help with time you to gain some mastery of image composition, color theory, and light – essential skills for an environmental artist.
These practices also help to acquire a certain artistic maturity, which makes it possible to be self-critical towards one’s work. To be self-critical is a good quality to have because every time you will need to readjust the environment or assets you create whether due to the art direction’s decision or your own reflexion.
All these qualities develop only with time and might be hard to gain. That’s why I suggest that you prioritize classical artistic learning which is longe rather than the technical side of 3D art. Learning 3D software, game engines and the workflow for building environments for games is something that you can learn way faster. These are just tools, like a paintbrush. Also, Art directors prefer to hire an artist with original works in the portfolio rather than artists with typical works that are everywhere.
I don’t have any course to recommend in particular, choose what will fit you. I personally decided to do a bachelor; for some people will find self-learning on the internet and taking online courses the best solution. However, note, that a diploma or a bachelor’s degree can considerably help you to get a job in another country. But in both cases, it’s important to continue learning. In the end, it’s your portfolio, personality and personal contacts that will allow you to land your first job in the industry.
Speaking about the contacts, it could be useful to invest time and money in the video game community: taking part in game jams and contests, visiting or volunteering for conventions like the MIGS, SIGGRAPH, IGDA and so on. Of course, doing classes is also a way to establish contacts!
For someone who starts making environments for video games, I recommend checking the articles below. They partly cover the things I spoke about and explain some aspect in detail:
- Creating Modular Game Art For Fast Level Design
- Investigation into Modular Design within Computer Games
If you want to have a solid introduction to texturing (texel density) and good baked normal map, make sure to check Leonardo Iezzi’s documentation Texel Density: All you need to know and No More Wrong Normal Map.
Finally, a part of my workflow for high poly rock sculpts was taken from Strkl’s polycount thread.