Hi Elliott, This is a great breakdown and very generous in sharing your process and insights, you came a long way from the vending machine days!
Are you planning on releasing the UE4 project to the public? Or only builds? I'd love to play around with it in the editor if possible!
A very detailed breakdown of the gory samurai diorama from a very talented 3d artist Frank Daniel Moen Vedvik.
My name is Frank Daniel Moen Vedvik, and I am a character CG generalist from just outside of Oslo, Norway! 3D has been a passion since my high-school days, and something that gradually transitioned from hobby to full-time endeavor over the span of six-seven years. I have no completed formal education, beyond an unfinished bachelor in Informatics, and a short stint at Noroff in Kristiansand.
I currently scrape a living doing small freelance gigs, and selling avatars and accessories in Second Life, while slowly piecing together my very first showreel, hoping to make it into the industry somehow!
This diorama is something I’ve been wanting to make for quite a while as a tribute to a manga by one of my favorite artists – Blade of the Immortal, by Hiroaki Samura. The scene is fairly loosely based on a showdown between the geisha Makie Otonotachibana and brothel henchmen, from the chapter Dreamfall – hence the title. I absolutely loved the sense of quick, decisive, violent motion those pages evoked and figured it would be a fun and an interesting challenge trying to make such a dramatic, chaotic composition work in a 3D environment.
Another important goal was to try to improve my production speed – the sheer amount of work needing to be done in this scene would always be forcing me not to obsess over and waste too much time on singular details – something I’ve struggled with a lot in the past.
I instantly decided on a few things – I wanted the final presentation to look like a classical statue composition, something akin to The Laocoon Group, with an untextured, smooth marble look. I also wanted to convey motion in a single frame, with flowing cloth and dismembered body parts clearly depicting a tale of a violent, fatal and instantaneous clash.
I also recognized my need for having more human characters, composition, and sculpting, in general, to put in my portfolio, and wanted to make a showreel piece that packed some visual wallop.
I will be the first to admit, I don’t have much in the way of an established pipeline. I love to experiment with new approaches, try different things, and outright wing things, just to see what works and what doesn’t. My workflow isn’t a hundred percent linear, something very apparent in my WIP-shots.
Generally, I tend to work on characters for real-time use, sculpting in T or A-pose and then progressing down the pipeline towards a low-poly rigged asset. This time around, however, I wanted to sculpt everything in pose, to make sure I didn’t skimp on the anatomy, and that every character, every limb, would look unique and interesting on its own. That, and sculpting tons of hands and feet were just plain
good practice with body parts I would often chicken out on!
The first thing I did, was quickly block out the bridge I’d position every character on, in Blender. I needed a foundation to plan out my composition on, and as the characters were my focus, I opted not to spend any significant amount of time on an environment.
I quickly decided that the main view angle ought to be from the side, and decided to remove one of the bridge rails to remove some visual clutter before I started blocking out my characters from zSphere stickmen.
In hindsight, sketching out the pose on paper first would probably have been a bit faster – but not by much. The important step here was to simply plan ahead, try to figure out roughly where and how each character would be positioned and sliced up. without getting bogged down by detail work.
By this point, I started to realize just how much work sculpting everything out would be, but I also found it exciting to see how quickly pieces were coming together – and how much fun it was to simply start working somewhere else in the composition whenever I got stuck, making me settle into a good sculpting rhythm.
After the first initial composition planning and anatomy block out, I went on to start working out the clothing. Clothing was added so early in the process, because I didn’t want to waste a lot of time sculpting anatomy details that clothing would hide anyways – clothing that would never change or be replaced! This is not a workflow I would’ve used otherwise. The cloth was made by simply masking out sections of my characters, extruding the section, and then using basic sculpting and moving brushes to give the cloth form. At this stage, the garments would look a bit “Blobby”, due to a low dynamesh setting – something I’d deal with in later passes, for now just focusing on the silhouettes and larger ripples of motion in the cloth. I also spent time at this point figuring what parts of the bodies would be visible, deleting the unnecessary sections, so I eventually could move on to chiseling in more refined anatomy.
I don’t really think there was much magic or finesse in my approach to the clothing – a faster way to go about it would probably have been to include Marvelous Designer in the process somehow.
I simply buckled down and kept hammering away at the details with simple, basic clay brushes, going over large sections in pass after pass to make sure the motion and direction of my folds, as well as the volume of the cloth, made sense. I experimented with the ClayPolish function to refine my shapes, and notably did a few passes with it towards the end of my process to get rid of unwanted noise and imperfections, as well as making the ridges and folds really crisp and readable, giving a pleasant, smooth finish. Keeping sections like the rolled-up sleeves and other pieces of layered fabric as separate sub tools also helped a lot with shaping folds and moving things around during the work process, as well as making working in tight crevices and complex, clunky spaces a lot easier – layers easily separated out meant I could break up tricky, interwoven shapes into individual parts.
I discovered a lot of useful things during this sculpting process. When I realized I could make organic, animated strips of cloth by lofting between curves using the CurveSurface brush, it sped up the blocking process a ton – and the IMM-Curve brush quickly became very useful for adding strips and layers of extra cloth in an intuitive and fast way!
One issue with a workflow like this, is that it becomes hard to use symmetry, as nothing is centered in the scene – except, in this case, the hero character. My solution ended up just blocking in placeholder heads for the males.
Later, I could make the actual high-detail head sculpts as centered, symmetrical, separate sub tools – easily welded into their respective bodies, later on in the process! I also did an asymmetrical sculpting pass on all heads after welding, as well as adding more fine detail, to make them seem more organic and visually interesting towards the end.
After the first initial, quick anatomy pass, I chopped up the bodies and pulled the parts apart – trying to imagine how the characters had been struck, and how the momentum would have made the parts rotate and spread out over time as they separated. The dynamics was a central aspect of the piece, to be conveyed with clear lines of action, and something I kept coming back to during every part of the sculpting process. I wanted the motion to be a part of the model everywhere – from the position and rotation of the airborne limbs to the waving folds of the garments, the ripples, and directional movement in the hair, and the streaks and spatters of blood. I constantly kept rotating my camera, shifting segments around, to make sure my use of negative space gave the scene a readable silhouette from all angles.
The hair on the rightmost assailant, as well as the female, initially proved hard to work on – a separate sub tool for each long strand of hair proving impractical and slow, while using dynamesh to weld the strands resulted in unmanageable, blobby shapes. I eventually discovered that Dynamesh had a setting for preserving polygroups as individual meshes – making it trivial to preserve the sharp crevices between the hairtufts, while easily giving the appearance of having welded the shapes together with trim and polish brushes. This has quickly become an incredibly useful tool that I use for a lot more than just hair now.
At this point, I had my first composition, having finished my first few passes of every part of the scene, I decided to put the model aside, noting all feedback and critique, while working on another project – until I finally came back to the diorama with fresh eyes, over a month later.
Getting raw, harsh and brutally honest critique is INCREDIBLY important – when you work on anything for long enough, you WILL eventually start getting tunnel sighted, and blatantly obvious flaws and issues with your work will become invisible to you. With fresh eyes and feedback in mind, I quickly realized the following issues with my scene;
The composition was too tight – notably, the silhouette of the heroine blending in with the man in the background. I spread out the composition a little more, while trying to keep the overall sense of flow in the piece a rough < shape. I had this suggested to me once more AFTER this pass, so I might not have pushed that aspect far enough – but done is done.
The cloth looked too liquid and wave-like, lacking the sharp creases and form-preserving folds and dynamics present in actual fabric – stemming from an insufficient use of references. Make sure you know how what you’re sculpting is SUPPOSED to look like, kids! There was no way around it – I ended up doing yet another sculpting pass on every piece of clothing, mainly using the dam brush for sharp indents and creases.
The far railing added too much visual noise, breaking up the silhouettes, and making the scene hard to read. I initially removed the thin bars from the railing to simplify the shape, but ended up simply stripping it off entirely, as well as the supportive beams, leaving just the middle walkway of the bridge – which really gave the characters some air to breathe!
I never wanted the violence and gore to be overpowering, refraining from modeling detailed guts and brains, as it’d distract from the things I wanted to draw the eye towards. However, at this point, most seemed to think the composition a bit lifeless and disjointed– and several recommended blood spatter to tie up the characters, as well as describe the motion of the heroine’s blade.
After these tweaks, the scene looked like this:
Initially, I had no idea how to go about the blood, so I looked around a little for ideas and different approaches until I came across these Youtube clips;
This gave me the idea to use curve-strips of zSpheres as “Droplets”, before using a very low-resolution dynamesh to weld nearby spheres into smooth, organic shapes.
This looks a bit crazy, but it was INCREDIBLY fast to slap together really cool shapes after some initial trial-and-error with sphere sizes and spacing – and after converting to dynamesh and a few subdivisions, produced this as a result;
The first few iterations of the blood had a bit much noise and too many scattered droplets, but I kept cleaning up the sphere base, until I was left with a pleasant, stylized mess of liquid streaks that read nicely, and conveyed a clear sense of motion and drama!
After setting up a simple composition in Blender, and waiting for my poor 670 GTX to spend a few days rendering out the final pass, I was left with a nice opening sequence for my showreel – as well as tons of new tricks to use in future projects!