LogIn3d artist Patryk Urbaniak talked about the production of the scene, inspired by Dishonored & Singularity.
3d artist Patryk Urbaniak talked about the production of game-inspired 3d scenes.
Intro
Hello, everyone! My name is Patryk Urbaniak and I am from Poland. I started to learn 3D in 2012. I was sitting in the theatre, watching Avengers when I decided I will give VFX a try. Currently, I work in Method Studios Vancouver as a Lighting/Lookdev artist. From my last projects that I can officially say, I worked on Guardians of the Galaxy Vol 2 and Spider-Man: Homecoming.
Singular Honor
In Singular Honor I was responsible for all of the visual aspects. I am a Generalist. So modeling, texturing, shading, lighting, comp, was all done by me. In the challenge organized by Hum3d we had to focus on character from a video game. I chose Corvo and Clockwork from Dishonored, but I also wanted to add a little something from other games that I really like – Singularity and Crash Bandicoot. In terms of modeling and texturing, I didn’t do any real humans before. Usually, I work on a proper lookdev while having a model and textures, to begin with. So going into semi-hard surface character was just easier for me. I decided to spend on this render maximum 80 hours. It was pretty hard to find that time while doing some other personal stuff and overtime.
The main idea was to create a fight scene. I wanted to make my main character look like he is very confident of his actions. Corvo knows exactly that he is going to send the Clockwork to different time. It is a cool spell in the game that you can bend time and use it in your advantage. Make things older or fix broken pieces. I started with simple composition, there is a page with lots of cool camera shots that can help you with that.
Production
Blocking the scene and preparing proper scale and proportions of boxes took about 1 day. In terms of software I used 3DS Max for most of the things. I also used ZBrush for weldings and finishing the mask. Mari for procedural textures, Maya for nCloth, V-Ray for lookdev, lighting and rendering. After Effects for compositing, Marvelous Designer for clothes and Photoshop for few fixes.
About the environment, I was watching gameplay videos and tried to make as many screenshots as I could. Also trailer was really helpful, bunch of references down there. I made a list of all the things that I was able to see and just started to close them one by one. In general, there is almost 30% more of the models we can’t actually see in render. After preparing the first composition, I didn’t know if I will use all of the models. So after some time, rearranging shot composition was a way to go, to make it more clean and readable.
Textures
Basically, there are only few textures in this scene. If we won’t count textures for books, there are only 7 bitmaps. Three grunge textures to mix with noises to break up the procedural look, two wood textures, one mask for the floor pattern and one leather-ish pattern for the glove. All other things like fabric, ceiling, walls, metal, chairs, etc are done procedurally. Also in terms of mapping, only few objects have a proper one, rest is just box mapped or unmapped, with TriplanarTex on it. I didn’t have to jump back and forth from Mari and Photoshop to 3DS Max. I kept it all in the material editor. It was really comfortable to work this way and render, because the render loading time took about 5 seconds for all of the things. Besides, if you are not using many textures, and the geo is modeled in a clever way, what is there to load? The main reason why I went into the procedural workflow is time. If you need a full list of features and maps from 3DS Max that are supported by V-Ray RT here is a link for you.
I didn’t want to spend few days on mapping and preparing proper textures of wood or breakups for surfaces to see all the details. When it comes to noises, you can just simply add one more level of detail and you already have it. No need to go 8k.
Lights
I wanted to make it as simple as I could with respect to light consequence. So I started with preparing geo that can actually be a source of light like chandelier and candles.
There are also 2 additional lights behind the camera, the one is from a door that was opened by Corvo before he entered to the room and the other one is just a simple warm fill light.
I am not fully happy with the lighting result. I feel like I am missing a lot of detail and story because of lighting. I should study references much more before I went into light blocking.
Render
I have used V-Ray RT for this one. It is amazing how optimized it is in 3.5 version. I wanted to test new features. The main advantage of GPU is how fast you can achieve results and that there is basically no settings to change. V-Ray RT gives you very good quality and optimized settings out of the box. You can do all the materials without turning off your render and it will update fairly quick. My GTX 1060 (notebook version) rendered this scene in 2 hours for 2000×1175 and it took 500MB of VRAM. While it took 7 hours on CPU (i7-6700HQ) and 14GB of RAM. I know what you are thinking, how the heck did it take 14GB of RAM if almost all of the textures were procedural. But like I said earlier, I wanted to test new features, so I baked all procedural textures into 4k, plugged them to TriplanarTex, just to see how it will work (on-demand textures). Changed the scale of the textures to match previous render and I made the test. About Adaptive light, the time went from 2.5 hours to 2 hours just by simply changing from full evaluation. The results were a little bit different, but I was able to fix that in comp. Next time you want to do some render with V-Ray, definitely check RT!
Challenges
The biggest problem was that I had too many ideas. But there is more. I started doing light before all of the aspects were clear to me, I also made a mistake of starting to fix things in comp to early. After few days it was pretty hard to overcome those problems. I should just go back to lighting, and to shot composition, by making it more dynamic. The main character is too static. I wanted to make him more like a human and just to give a little bit more life to him but unfortunately Im not an animator. I didn’t know how to do it properly. Also my shot composition just killed all the models and shaders I invested most time in. I would like to go back to this scene and give it a try again.
About the production tips. Take your time and prepare a really good base for the shot composition. Think about a render like about a photograph. Remember that lighting can change the appearance of every shot, making it agressive, calm or interesting. Keep in mind that adding movement will help your render, even if the frame is static. Add life, proper atmosphere and some story to your renders, it will give you an awesome boost to the final result.
Thank you very much for having me here, and for reading. Take care!