Erasmus Brosdau talked about his new amazing project, which takes real-time animation production techniques on a whole new level.
A lot of things happened and are happening, the industry is moving fast and everything gets more and more tailored towards real-time rendering. I gave a lot of talks around the world showcasing the work with the Lord Inquisitor project, which opened new doors. Professionally, I now work as Senior Art Director for AMD now, where I work on various real-time demo projects that we show around the world, mostly using Unreal Engine 4. The release of the Prologue of the Lord Inquisitor marks the end of this production sadly, it’s a decision I made after it was complete. All the work that went into making it (including around 5 years of learning and redoing), made me feel really exhausted.
Same was true for my great friends who helped to create it, it was like a free-time hobby project turned into a professional production with actual deadlines and super high quality. Having an 8-hour workday before that with “real” deadlines, this can quickly grow over your head. At that point we decided we can only go on with a professional budget and production committee to work full time on it, otherwise, we would never be able to satisfy the demand in quality and runtime. Besides that, as we are being a fan production, the bigger your project gets, the more problems you have with the original IP holder. More and more problems occurred and thus the decision came along to only continue professionally with an official license, or to stop the project.
That’s the situation we are in right now and there is no other choice pretty much – without an official license, we keep on walking on very thin ice, same fates as other big fan projects of other IPs so to say.
As we didn’t lose our passion, however, we created a new project that can run in a different media format (web series) and which is not tied to any popular IP at all, which makes this new production so much more relaxed and also removes all other problems about licensing. I was really happy to see, that most of our fans could understand this decision and really enjoyed our new production called “Origin Zero”.
I really like creating new worlds and stories, so I always had some ideas in mind. When I started designing this universe, I knew I could never beat Warhammer 40K in its epicness and I also wanted to avoid any problems in terms of making the appearance that I now just rip off that franchise and claim it to be my own one. So I intentionally moved away from anything Warhammer-related but wanted to stay in the sci-fi universe. I have an established style already in my designs, which you can also see in the IP called “VR Score” or “Codename Skyharbor”, which I designed for Crytek. That allowed me to create things much faster and most designs were created in only one day, except the suit and helmet designs.
I always want my designs to stand out a little bit from the current sci-fi trends, so it took me a bit to find some unusual designs, that still look cool and also functional. Once you animate Space Marines, you quickly notice that their suits are actually their biggest enemy and makes it impossible to move. I work on all my designs by keeping the animation factor in mind and also often make a simple rig in 3D to test the range of motion. I didn’t have any references, I just designed a universe that I wanted to feel slightly unique and that had an awesome backstory. That one actually took me a long time and was the first thing I thought of, before doing any design work. I’m very proud of the core story, but with the first episodes we pretty much can’t show any cool twists or something like this. In fact, Episode 01 is supposed to give a lot questions, instead of answers. Episode 03 is the longest with already 10 minutes and gives a much better insight into the universe. We already shot motion capture for this one and it might release end of 2018, but somewhen in 2019 is more likely.
The whole cinematic is a one-shot camera track and you could imagine it a little bit like a landscape corridor. At least, that “corridor” area is the most detailed one. It’s a regular square terrain shape with lots of trees in it and just the area the characters are walking through is detailed and designed to fit the camera nicely. As we’re having 8 characters and 2 robots in the scene, they all change their position once they exit the frame. Once the camera shows them again, they all appear in a nice framing. Sometimes also I had to solve animation cuts as we didn’t have nice looping animations, so in one specific section one character comes in the foreground and blocks the view of another one. In exactly that time that blocked character jumps back to the first frame of its animation. As the camera moves on we see the character showing up behind the foreground character again and it looks as if nothing has happened. When you watch the cinematic from a birds perspective you can see all characters moving and rotating around the forest scene as the camera moves between them, it actually looks kinda funny.
For the start of this cinematic, I knew I need a nice looking forest and one big metallic tower that came through the ground for not yet explained reasons. In my first previz of the cinematic, the tower was just a simple cylinder. I then turned to a kitbash approach and actually reused a couple of sci-fi skyscrapers I had for The Lord Inquisitor. I also threw some old spaceship inside there, detached elements and put it on positions that looked like they belong there. To make the mechanical tower took just a couple of hours and I also didn’t make a concept for that. I just wanted it to have larger flat areas with some sharp edges and bigger cables coming out in various sections. Luckily, it looked quite nice in the end result, but the tower only works for this perspective.
I’m not good at vegetation modeling and I also wanted to focus on the overall cinematic, not on modeling trees for weeks. So I thought I could buy some great tree assets from the UE marketplace and pretty much call it done, but that didn’t work out at all sadly. I started with some nice asset pack of the redwood forest, but it was set up in a ridiculous way that cost so much performance and actually didn’t look that good. Besides that, I felt that the forest needs to be much more mystical you could say, lots of winding old trees and overgrown vegetation. Also, I wanted some big roots here and there giving them a slightly dangerous and spiky feel. I found some cheap model of a Chinese Banyan tree which pretty much had a look I liked. It was looking very bad and was very low poly, so I completely reworked the textures and geometry and assigned vertex colors on the trees so they could move a bit. I did the same with an old wooden trunk that looks like an overgrown arc. For the whole forest, I only used this one tree and the one arc-shaped trunk.
Different scale and rotation do the magic there. All other background trees and bushes came from that other scene from the marketplace and I did some retexturing on all the leave textures, as well as the ground materials, which I replaced entirely using Megascans textures. It turned into much more work than I thought, but the end result looks quite nice, although I think a proper vegetation artist could make it a lot better. I even made a simulation of stones in 3Ds max, once the soldier, in the end, gets dragged away, just to have some interaction with the ground. It’s a subtle effect but helps to make this look and feel much more realistic. Realism is actually something I don’t follow too close with Origin Zero. Everything has a realistic look at the materials to it, but the main goal is to create appealing images. This also is visible in the designs of the suits and robots, while they are functioning correctly to the biggest extent, there is a small fantasy touch to it that gives it instantly away that it’s not supposed to be serious realism. I also chose very vibrant colors instead of the typical desaturated sci-fi look. In episode 02 we show human faces and you will also see that they are intentionally slightly stylized. I wanted to have some nice Global Illumination in the forest but didn’t go for a baked solution. Instead, I was using the Light Propagation Volumes that gives some nice results when enabled. Sadly, they are not being updated by EPIC anymore and are kind of a hack to activate. However, together with Distance Field Ambient Occlusion in a skylight set to movable, I was able to make some really nice looking lighting with great shadows underneath the trees.
Like all other visual elements, I had to make this somehow on my own. Just like the tree assets, I thought I could just buy something from the marketplace, put it in and call it done – but it didn’t work again. The hardest asset in real-time VFX are of course the sprite textures, so I was able to find some really nice looking ones in the marketplace that gave me a kickstart. The actual effect was then however very badly animated and setup, so I had to sit down and learn the particle editor for UE, which I hoped could be avoided. I at least understood the basic functionality and while the editor feels very outdated from its design I was able to adjust all effects as I wanted with colors and look. The heavy usage of volumetric fog gave problems to the transparency rendering and I had to find ways to make them fit inside this atmosphere. I worked on all VFX stuff for this demo and they actually gave me a lot of headaches. So, with a couple of assets from the marketplace, I was able to make them look the way I wanted and didn’t need to start from scratch, this definitely was a massive help. I’d love to work with someone being a pro in Houdini at some point as it can produce amazing results. I don’t see myself at all as VFX artist though, so I don’t invest in learning it.
I’m very routined in creating hard surface assets and nicely looking materials, so I didn’t need to invest in look development first. However, I changed the way I work with these types of assets dramatically. None of these assets have ever seen ZBrush, it’s only 3ds max and Substance Painter. I thought of a new pipeline for making these assets faster as I only want to focus on the direction and design. Making assets doesn’t give me much pleasure anymore, I see it only as a necessary task in order to be able to tell my story. Of course it’s still fun to work on cool assets, but once you look at your list and you see you have to model dozens of crazy robots, suits, weapons, ships, etc – you start to think about how to speed this up. The soldiers and the robots are having the same color palette as they belong to the same team. As I had exported a couple of textures and was working on the robots later, I noticed that the blue tone I chose back then wasn’t really working anymore. As a result, I had to go into all substance files and reexport like crazy a new blue tone I had chosen. I had so many assets with a unique texture that defined that blue tone and that was the point where I said goodbye to unique texture creation. What I then did was to use Substance Painter to only export ID masks and blend all the materials within UE4. There now only is one material that defines this blue metallic paint and its being used on all assets. When I want the blue to be slightly different, I just change the main material and all assets will have a new blue tone. Normal maps were directly painted in Substance Painter, I never baked any detail normals from a sculpt. The only baking process happens at the very start of the texturing process, where I bake the normals from a nicely smoothed version of my low poly asset to the same low poly asset having only 1 smoothing group – that saves a lot of draw calls and results in perfect shading. UE 4.19 now has another system for material blending and it enhances the system I was developing for Origin Zero. So now, I have to remake all materials again sadly, but it’s very beneficial in the long run.
The workflow here was very standard I would say. The rigs and robot animations all came from my good friend Soheyl Mohammad. I work with him for many years already and we became a great team. He is very amazing at animation as well as rigging, so that helped a lot. He did create the original rigs for all assets and hand animated the robots. The human body animations came from a mocap shooting done by metricminds, a company in Frankfurt am Main, where I live as well. We had completed shooting the animations for Episode 03, but I surprised them by email saying I will make two more smaller episodes before that. I showed them my very bad looking previz and they understood what I was looking for, jumped into a suit and recorded generic walking, shooting, falling and other animations. I then picked the takes I liked most and received them cleaned. Soheyl then added tweaks to them like hand animations, posture adjustments and things like that. This got retargeted to the rigs he made and then exported via FBX into UE4. I then just moved and edited the animation clips like crazy in order to make everything fit the camera and vice versa. Soheyl gave me a walking cycle for the big robot, so this animation is looped and I just animated the walking distance manually. So I think that was a great team effort here, especially from the mocap and keyframe animation side, but nothing fancy in the actual workflow.
I always wanted to do a one-shot cinematic and also had it planned for the final balcony scene in The Lord Inquisitor prologue. It didn’t work there at all, so I had to remove it. Once I was designing Origin Zero and thinking about the episodes, I started with Episode 03, the very long and complex one. Then one night I had a dream and was exactly seeing episode 01. When I woke up I thought, man, that could be a really cool first intro and I can finally make my one-shot cinematic. I then sat down and made the previz for the episode in order to visualize my idea to the other guys. Then we continued polishing it and I just changed it slightly in various areas to improve it, but in general this first episode is exactly like the dream I had about it, which was very funny as I had never worked like this before – might also be hard to control for future productions. I’m a big fan of Metal Gear Solid V and it only has one-shot cinematics. It’s obvious that in some scenes it works nicely, in others not so much. You can’t really use one-shot cinematics in every situation as it can result in a lot of problems, for example how to get from point A to point B. Besides that, you mostly lose a lot of connection to the main protagonist as the camera always have to move around to show what’s important next. This, however, fits perfectly for Episode 01 as the soldiers are not interesting at all from a character perspective.
The actual main character is the creature we don’t see. The soldiers are a symbol of power and control and because they all die within a couple of seconds, it showcases the strength of the enemy. So it didn’t matter who these soldiers were, I wanted to give more the feeling of standing just in-between them and not being able to see the enemy attacking you from somewhere in the forest. In UE4’s sequencer, I animated the entire camera by hand using tons of keyframes. I started with only the most essential keyframes to see the pacing and route of the camera. Then I placed my soldiers and other elements to see if the speed of the camera matches the soldiers, etc. The cameraman is not existent, so I can speed him up to travel bigger distances and also move the camera through small areas without the soldiers taking notice of that. At some point, I had to lock the animation of the camera which then allowed me to fine tune it. As the overall path and movement were in place, I added subtle camera changes and also made some up and down movements to make it feel more like a shaky hand camera. On top of that I made a blueprint that adds procedural noise towards the animation, enhancing the hand camera feel even more. I found that to be pretty complicated and hope the sequencer get a better solution for camera shake in the future. It was interesting to see that first person camera is a love/hate thing, while almost everyone really liked my designs and the first episode, many were complaining about the camera and that they don’t like it at all. The other half, however, especially pointed out the amazing camera work in this episode saying its brilliant and really unique. So, I knew a little bit in advance that this would result in a lot of negative feedback, but to me, that’s not a big problem as this one-shot camera will be only used in Episode 01. All other episodes will feature “regular” camera work and cuts, not because of people like that more, but because it serves the purpose of telling a story much better.
Episode 02 is already in the making and has a similar runtime like the first one. From its content, it will be quite different though and features some new challenges we need to figure out how to solve them. I made a traditional storyboard this time, all painted in Photoshop so that the animators know what needs to be worked on. This is all complete and I now work on the white-box models of the characters so that we can get the first animation passes done. I hope that this new episode will be live in around 3 months, of course, it will be uploaded exclusively on our YouTube Channel, which you can find here,