Prozessstills und Lookdevelopment: Viewport, Rig, Curve-Editor und Materialtests. Von feuchten Highlights bis zu matten, porösen Flächen - der Weg bleibt sichtbar, auch wenn einzelne Schritte unvollständig wirken.
Hier landen Screenshots, Clay-Renders und Zwischenstände, die im Visual Concept keinen Platz haben.
Disclaimer: There are definitely better and probably more logical ways to solve certain technical problems, but yeah… it was my first time.
So I guess the most fundamental part of the process was my intention to do everything inside Blender. This worked more or less well, but texture painting was something I had to do in Substance Painter-Blender just wasn't fun in that regard. I originally wanted to create this creature as a test and built the base mesh in 1.5 hours, which is ridiculous compared to the time I spent on every other stage. Then I started retopologizing the mesh by hand for roughly 8 hours and managed to do it somewhat fine. After that, I base-rigged it and went through basic texturing-I think UV mapping it, finding the whole visual language, and setting up the lighting for these scenes took roughly 50 hours in total. Only after that did I start the animation.
Within Blender, I used the FLIP Fluids addon for the water simulation. Understanding it was quite interesting, to say the least, but in the end, it really served its purpose to create great-looking and interacting water (if you ignore the repelling waves from the domain boundaries, lol).
For the camera, I used the Lens Sim plugin by Håvard Dalen, which is absolutely amazing! The camera results and the small nuances that the different physical lenses provided were fascinating to work with.
I had never properly rigged any object in Blender before, so I guess I was a little overconfident and unrealistic about the impact it would have on the whole project-especially how complicated it was going to be. Basic human rigs already exist; you can basically get a perfectly rigged model with optimized constraints right from the start. Anything that goes beyond repeating standard anatomy needs to be built from scratch. I feel like I chose the most annoying anatomy and complicated rig to start with, because it needed to be adaptable to three main requirements: posing, animation, and natural movement. Since my chosen creature isn't really supposed to have rigid joints compared to humans-who have pretty logical joint constraints-the whole thing required a complex setup to work.
The IK (Inverse Kinematics) constraint could technically give me the perfect posing for every limb, head, tail, etc., which is ideal for the baseline setup. But when it came to actual animation, it struggled so much to achieve proper movement and make it feel natural. I worked on 15 seconds of animation for over 20 hours, and in the end, it still felt stiff, rough, and illogical because it wasn't organic. The movement of the body didn't have an impact on the tail or rest of the body, which just didn't feel right. You need to feel the forces the body creates and automate them so the rig can almost move on its own; the bones and constraints need to react dynamically to the movement of the rest.
The only viable option for that in Blender was the Damped Track constraint. Damped Track allows for completely natural movement, but it always snaps back to your reset pose-so you aren't able to pose the object manually at all. It creates perfect movement if you tune it right, but it completely lacks the ability to work within a keyframed narrative.
I had literally no idea what the perfect solution was, because if you try to combine both systems on the same armature, you get feedback loops. Take a tail as an example: when you combine both, you end up trying to control the tip of the tail and the root of the tail at the same time-which just doesn't work… I tried.
My final solution was ridiculously complicated to manage with my knowledge level, but somehow it ended up working to this extent:
Almost every moving limb, tail, head, etc., works with 3 parallel bones at the same time:
A Control Bone which directly drives the mesh.
A Damped Track bone chain for dynamic movement.
An IK bone chain for precise posing.
The Control Chain uses a Copy Transforms constraint for both of the other chains to blend their influence. Afterwards, I created a custom properties driver system which manages the influence of all Copy Transforms with a single slider in one clean interface. These drivers are animatable, so I could keyframe myself from a posed position to a damped one, more or less seamlessly. But since it's still calculated dynamically, you still get bones drifting apart caused by feedback loops-though it remained manageable.
Horrible… I really hope there is a better solution for this, but I couldn't solve it any other way.
I guess the most frustrating part for me was the storyboarding and imagining what was going to happen in the scenes. At this point, I had already redone several animations multiple times just to fix the armature and how it was working. This took so much energy that I really struggled to find the narrative and the core idea-which shows in the final result (more on that in the conclusion). I worked with Non-Linear Animation (NLA) to create my storyboard within Blender. In the end, the rig needed to be baked into keyframes so it could interact with the FLIP Fluids simulation, which also needed to be baked.
Afterwards, I did the sound design with Vincent Müller and got help for the edit from Emilian Tsubaki to wrap it up. I then finalized the remaining sound design by myself-which you can probably tell from listening to it, I suppose.
Conclusion: It was overwhelmingly complicated to do all of this by myself. It took a ridiculous amount of time to finalize, and the process really dragged on the harder the problems were to solve. I would definitely change up the whole pipeline if I were ever to do this again. But it was also fascinating to dig so deep into solving these issues. I feel like the final product could be better in so many ways, but I got really close to creating the impression of something feeling real-and I'm super proud to be able to achieve something like this through actual 3D craft without generating a video through AI.
