Geoffrey Rosin talked about the amazing things he’s done during the production of assets for Dishonored 2, including the ’Stilton’s Manor’ level assets.
Geoffrey Rosin talked about the amazing things he’s done during the production of assets for Dishonored 2. In particular he worked on the amazing ‘Stilton’s Manor’ level! 80.lv must admit that Geoffrey’s work is a bit of a miracle. He’s elevating the traditional elements of 3d games to some extraordinary levels of artistic talent, helping you getting lost in your virtual worlds. One of the best artists in the immersive sim genre. Thank you so much for this interview.
Hello! My name is Geoffrey Rosin, I’m a 26 years environment artist and I started CG at 13 thanks to my father. We were playing games like MoHAA (Medal of Honor Allied Assault) in LAN and very often he was getting custom maps online. So I decided to learn to use MoHRadiant Editor bundled with the game to create my own levels.
When Half-life 2 got released, I learned to use Hammer on Source with a Level design french community called “Mapping area”. Later, I got invited to do modding projects such as “I-robot” , “Get a Life”, “Bisounours Party”.
I also worked on E.Y.E. Divine Cybermancy, which was designed to be a free mod at first but later evolved into a full size game with AAA ambitions and was the foundation of Streum on Studio. It was my very first
serious project during High school; I was mainly doing level building and level design for E.Y.E. but my main concern was about visuals and I started to feel frustrated to not be able to get the models and textures I wanted as I was imaginating them.
I then decided to enter at H.E.A.J. (Haute Ecole Albert Jacquard ) , a Belgian CG school, to learn modeling and texturing. During my last study year, I was working both for School and as Freelance to maximize my chances of quickly finding a job at the end ; Thanks to a friend, I contributed to GameTextures.com and their awesome material library.
During the same year, I met Shawn “FMPONE” Snelling, a Counter-strike map maker with whom we did Gwalior, a CS:GO level which got officially added in Operation Bravo. It allowed me to get in touch with New World Interactive and work on Insurgency as an Environment Artist. However, I also just heard back from my application at Arkane studios and decided to go for it. Now, I’ve been Environment artist at Arkane since August 2012 and I had the pleasure to work on “Dishonored 2”, its DLC and helped on “Prey”. I’m also regularly teaching GameArt in various schools and still doing personal project on Source engine.
At Arkane, for the most part, Level Architects or Concept Artists will often provide a preliminary 3D mesh used for visual intention. It then needs to be detailed in Highpoly, cleaned and optimized for the low poly, baked and textured. Since design questions are already answered, we just have to make it functional by adding screw, weld, hinges and such. The same process goes for everything (lamp, furniture, wooden statue, building, etc). The process is quite straightforward and It’s during the texturing phase that we often get the most creative freedom.
When starting on a model, I first sync with the guy who provides the mesh (or concept) to make sure I got his initial intentions. I can also suggest ideas, improvement and modifications (visible on the previous screen, on the glass part) while keeping player navigation and gameplay in mind.
Since I especially love texturing work, I absolutely want to collect info and be inspired as much as I can before the first stroke. Defining what type of plaster I will chose, what Screw shapes, stains colors, etc… Everything that can make your asset feel original and surprising is good to take. Also, don’t be afraid to mix them; you can blend the shape of an object with the material from a second one and adding micro details from a third one.
In a way, every details matters when you love your craft. But in production constraints, Time is always a limit. Also, consider if the details are going to be seen by the player. I think you can always use quick old classic texture recipes on hidden parts to save time since nobody will care except you!
Story of an object
Environmental storytelling is always important and making an aged leathery chair showing signs of use rather than a brand new one can contribute a lot to the mood of a scene. The asset’s “visual story” can be brought both in the modeling and texturing, even the way the player will interact with it. By being subtle, you can explain with accuracy how nails falls off the chair, how moisture weakens fabrics and can lead to springs poping out the cushion.
(In this case, I did Dust decals to mark out the previous painting places, add ill colors and dirt on the Wallpaper and teared up its edge to get interesting details. Thanks to additional meshs, I was able to make it pop from the wall and show the previous wallpaper covered by the new one. )
But you always need to evaluate if these details will really be essential if your assets is to be placed in a dark corner of a room, or hidden by other assets. Sometime, Adding too much story and detail on the asset can feel heavy and create visual noise which may become an issue once it’s placed in a fully set dressed environment. So you sometime have to let go a bit and favor a modest layer of generic dirt.
However, If the goal of the object is to impress the player by having a crucial role in the theme, mood or composition of the scene (Hero props), then it deserves that you spend some days on it. Of course, reality is never that binary… Sometimes you have to take in account how big is the object on the screen even if it’s a generic asset, such as a building facade. In that case, it obviously also merits your complete and deepest love
As said previously, we’re not in charge of a precise level and its lighting. We get an object, often already designed and modeled and we add some texture magic on. Nowadays texturing is getting easier and easier with new software like Substance. If you want your work to be unique, you’ll principally have to find ideas, not techniques. Your goal is to find the little details, the right story to tell into your texture.
(Duke palace Floating lamp…)
For the huge Spotlight I did for Dishonored 2, which was supposed to repel bloodflies, I added some dead flies inside the glass and simulated their thickness by adding gradients. The dust covering the molted glass also helped to get a nice story. Where the metallic bars disappears in the glass, I made them occluding the light from inside and appearing like shadows. All that directly hand-painted in the Emissive map.
(The “Rat Light”. In Dishonored 1, it was made to afraid rats in street and the idea was kept for Bloodflies… but we kept internally calling it a “Rat light” instead of Spotlight ! )
Regarding How I work with my different maps, let’s say that I just avoid telling the same things in each channels and It allows me to get nice effects with a wide range of lighting setups. I put most of my time into my Albedo and gloss for that reason, because of the way it’ll be lit in game;
(Curator building’s stones)
Albedo and gloss work tends to stay noticeable in ambient lit area. I would like to insist on “In game” since lot of CG students I had were always checking their work in Marmoset (which is a really great tool for asset previewing as a personal satisfaction) instead of the concerned level.
(The curator building // Design by Valentin Levilain)
There are cases where you won’t be able to tweak the lighting the way you want to make your textures (or 3d asset) as cool as they should be because of Gameplay constraints, scenery readability, perf budget, it’s obviously better to directly adapt your stuff for the game you’re working on rather than the contrary.
(A broken and dirty marble floor did for the Dust district mansion, The Bunker )
When comes the time to add details, since Sebastien Mitton (our art director) wants the game to feel like a painting, we pay attention to volumes and masses in our textures.
Each existing materials can be synthesized into different frequencies (I prefer to talk about Frequencies rather than noise ) and this is what Substance Designer permit you to toy with. Dishonored stylization can be considered like a tweaking of these frequencies as we generally wanted less high frequencies (for instance: small numerous dots).
For instance, the way I’ll add dust on a wooden floor will be different according to the wood pattern complexity to avoid making my texture too complicated.
I prefer to use flat colors for smaller details (around 3 pixels). When they get bigger, I start looking for the right frequency to add as the simplicity betrays the detail’s understanding. Honestly, with any uniform color and the matching gloss value, you should be able to do all small details you want (Wood pokes, metal scratch, moss, etc).
As said before, I did lot of hand painted work in my textures but it never prevents me from using photos to directly get the frequencies I was looking for, since we did 95% of the game with photoshop.
Working on materials
For Dishonored 2, I used Photoshop for the whole production. Substance Designer and Painter were at their very beginning when Allegorithmic came to present us their tools at the start of production. We felt it was a bit late to completely change our workflow. It’s something you spend time defining during pre- production and consider locked afterward. That would also mean to adapt export tools, workspace to get a rendering that match the one you have In engine, etc. So, we only shifted toward Substance at the very end of Dishonored 2.
Personally I’m a huge fan of Designer since I was working the same way in Photoshop. I still make my height maps in Zbrush but do the detail and color passes in Designer. When I was working in Photoshop, I didn’t have to think about how to do broken stones of a wall for example. I would sculpt them directly in Zbrush so I could eventually reuse them as a unique mesh later.
It’s something I found tricky at first to setup in Substance, I preferred to stick with Zbrush for this step (Nothing exceptional, just the classic workflow : Sculpting Few blocks and mounting the wall). I feel I’m still able to iterate easily and quickly exporting both an ID and Height map from the sculpt.
I’m absolutely not saying it’s better to do it that way. I mean, according to the time spent, the result I get isn’t necessary better than what some people are able to achieve by only using Designer. But with this workflow, I can reuse my sculpt to get directly the 3D meshes and it’s really useful for derelict and broken stuff.
It depends on the engine or shader available. Some engine takes a BnW emissive mask that’ll brighten up your diffuse when others will use a new texture brighted up by an emissive value and this is what we currently use in Void Engine (Arkane’s custom engine).
The best way to start is to know which kind of material are you working on. Only your ideas matters. Your goals is to avoid throwing a flat white texture and call it done. Think about how to simulate the way light is the brighter under the glass layer. Since light spreads inside the glass (refraction), you can safely go for a big gradient for a start.
Then I’d tweak the color in a new layer. For this step, I really like to choose a punchy color that will contrast with the one I used for the diffuse. For instance, I added lot of cyan and green for the Ratlight emissive general look. Using the initial gradient, I don’t shy away from saturating its darkest parts and boosting up the brighter ones. It’ll add more depth to the texture.
The next step depends of the Light size ; For smaller objects, I would often finish with a very subtle medium frequency dirt layer since the prop itself brings enough frequencies into the image. For bigger one (Duke Floating lamp), since the lighting part is able to cover a large part of the screen, It’s better to make the texture more appealing by telling more things.
Take care of the position of the elements you add, if they’re “on top of” or “behind” the glass. In the second case you shouldn’t be able to see it clearly because of backlight effect. To simulate depth and light propagation, you can vary the size and the position of your dirt layer (Dead flies, dust splat, etc ) inside the light. You’ll will see its edges getting blurrier or sharper (when very close to the glass).
For a dust covered lamp, a slightly darkened copy of the original layer added on the emissive will be enough. Since the Dust layer is thin, you should get some light passing through so avoid setting it totally black.
To sum things up, think about how the different parts react and find a cool way to synthetize it in your texture. It’s often boils down to gradient and colors. Also keep in mind the temperature you want (Hot sensation on touch, etc) and adjust your values.
All these methods are pretty old school and might differ a bit with most advanced engine, but I think they are part of the fundamentals.I’d love to be able to play with parallax effect ( Bump offset in UDK) thanks to a glow texture to simulate better the depth of the light for example. Maybe I would also corrugate the glass to get some refraction effects. The way it react depending the view angle, how it reveals its details, I think all these would makes a lamp more interesting to discover. With a simple emissive, it stays limited!
How does the light help you build better presence and better atmosphere in the scene?
Hmm, it’s not easy to answer since it really depends of your goal. You wont build lighting for a game the same way you would do for a personal project since gameplay needs to be considered. For instance, in a game like Counter Strike, it would be a bad idea to let some areas in deep shadow as player would be able to hide and makes the competition really unfair.
Low contrasts lighting are the easiest way to get a good result with such gameplay but it can be super tricky to get quality beauty corner with such constraints. In the major part of game, light is often used as a tool to drive the player attention to a story element, a puzzle, the exit…
Obviously, each light doesn’t have to serve such a purpose and can be simply used… to lit the scene, in the artistic sense. As for textures, light can be worked as volumes since it makes a shape in the scenery. You can see it the same way as in paintings, the way a focus point is highliteghted, how it drives the lecture of the image. But in a case where the camera is able to move everywhere, you cannot expect your scenery to be perfect from every angle. Having dynamic or baked lighting can also make a huge difference: Not the same look, not the same performance cost ect, as it was the case for Dishonored 2 compared to the first one.