Labcoats and Lairs

Based on my JelloApocalypse Danganronpa animatic

(@jelloapocalypse @chulacabra @insertdisc5 @eternal-savvy-blog. I actually don't know if I should ping you for that since it's a making-of of a fanart and not a fanart, but at least it gives me an opportunity to thank you all again. Also don't feel obligated to read this 2001 monolith of a post, this is mainly in the interest of sharing knowledge)

So, in case you stumbled upon this post first, that video was based on a running bit developed by Jello and his friends during their fully-voiced playthroughs of Danganronpa 1 and Danganronpa 2. This is definitely the most ambitious piece of fanart I ever did, though not the first of its kind (In 2021, I did a similar project for Overly Sarcastic Productions), and I want to talk about how it was made in detail, because I feel like there's a lot that could help other growing artists.

For example, here's one of the most important things to know: It took two months to make, but I only started outputting shots about a month prior to the final stream, in early July. Before that were several weeks of pre-production, learning Blender, creating the 3D environments, and building a solid workflow to make my life easier down the line. This post will focus on that first month and be more about sharing tips and advice than fun facts, so if that's what you're interested in, stay tuned for part two where I'll go over the whole thing sequence by sequence.

1- The idea

The backstory behind this one probably won't surprise you. You listen to a song a bunch of times, it generates cool visuals in your brain featuring your favourite characters, and then this awesome music video either lives rent free in your brain forever or you laboriously turn it into reality somehow. Fortunately for me, I had to learn Blender and add stuff to my demoreel, so that way I could channel the ADHD into something productive. Because it certainly wasn't any good at getting me to follow tutorials.

Which is actually the first tip I can give to the artists in learning out there. If you want to know how to use a certain software or get better at a certain skill, but practising just for the sake of learning isn't doing it for you, give yourself a medium to long-term project and learn what you need as you go along. It's easier if it's in service of something tangible. Even better if you know that by the end, it will make someone happy, for example a Youtuber who you know regularly sees and likes fanarts. You'll definitely need that added motivation.

2- Start rough to get the dough

So I got to it and began doing a rough animatic. Which, if you want to make a project like this, I'd recommend not skipping on. I know your idea is perfectly clear and punchy in your head, and you need to hold onto that. However, there will be issues you can only see with your actual eyes, so might as well snipe them early so they won't waste your time later. And you can phone it in as much as you like. Who cares if it's messy and incomprehensible to anyone but you, it will still give you a solid base to figure out the editing, timing and continuity (look up the Knives Out storyboard).

I made it with Procreate Dreams and imported it into DaVinci Resolve. I then broke it down into five sequences and numbered the shots by increments of ten, in case I needed to add shots in-between without messing the order, with a frame counter for each (DM me if you want to know how to do that in DaVinci). If you're curious, the final total was 84 shots and 4972 frames.

And lo, the idea had a physical form, which meant it was time to move on to phase 3.

3- Meeting hot new 3D softwares in your area

Now, I wasn't an absolute newcomer to Blender. I had done the donut a few years ago, I animated a walk cycle with it, and I knew Maya like the back of my hand, which is similar on many fronts. But I can't say I knew Blender like I do now. It's not just about the individual tools, windows and keyboard shortcuts, it's about understanding its design philosophy, understanding how each different part relates to each other, understanding the way it interprets data, until you can instinctively grasp each new function you come across because they make sense with what you know of the software. And because that last sentence was very pretentious, here's a funny image from a shot in WIP stage.

I started by building Kyoko's office, making all of the 3D objects myself. I could have grabbed some of them on an online marketplace, but I really wanted to add that "everything was done by me from scratch" in the credits. Plus, the point was to learn.

At first I kinda struggled with Blender's modelling tools, occasionally going back to my comfort zone (Maya), creating and exporting the models from there. But once I had a good grip, explored the features, read the documentation, set up a nice workflow, learned the keyboard shortcuts and added a few of my own, it was mostly smooth sailing. And I stuck to the most basic props and furniture imaginable, which luckily fit well with the sterile and alienating work environment.

By the way, that's another very important tip. Make your life easier. If there's friction, your first instinct should be to stop what you're doing and look for a way to alleviate it. If there's a repetitive task or action you have to do over and over again, there's probably a way to automate it. Learn the key bindings. Alt+tab at the first sign of tedium and look for an alternative in the Blender documentation. Forage the forums. Peruse the plethora of add-ons. Find the video with 300 views from the Indian guy who has the perfect solution to your very specific problem. I am formally giving you the mission to develop a taste for optimisation and getting the best effort/result ratio. Making detours to find shortcuts may feel like a waste of time at first, but spending hours brute forcing through an easily solvable annoyance is like trying to swim straight across a river because the bridge was ten meters away and you couldn't be bothered.

Also, this includes having a strict naming convention for your objects and a well-organised folder system, both in and out of Blender.

For the latter, I went off the default Maya folder system template, slightly adapted to suit my needs: "scenes" for the individual shots, "assets" for the files containing the elements I reused in all of these shots, "movies" for the mp4 files I used for editing, "sound" for the audio clips and "misc" for the sticks, ladders and minecarts.

So I now had a set, in which I could place my camera anywhere, but I needed it to look good as a background for my 2D characters. Fortunately, that's when I discovered Blender had an integrated compositor, and that made a big difference.

4- Isn't compositing how you get feritilizer?

If you're not familiar, compositing is often the final step in video creation. It refers to the action of combining different images and pieces of raw footage to create a single harmonious shot (think green screens and VFX integration), though it generally encompasses everything that has to do with polishing an image, like colour-grading or 2D effects.

And when it comes to CGI, this "combination" process is on a whole other level. Because a single render from a 3D engine is actually the sum of many individual images called render passes (or AOVs), calculated independently and then merged together. One contains only the direct light, another only the indirect light, and another only the refracted light, to cite but a few. All of them can be isolated by a compositing artist to tweak them individually.

But if there's more than one light source in the scene, you can also render those separately if, say, you want to slightly alter the hue of the bedside lamp without affecting the light coming from the window. So you multiply the aforementioned render passes (direct light, indirect light...) with the number of light sources. And then you do it again with the foreground, background and middle ground, all rendered separately, and for a final touch, you sprinkle them with utilities AOVs, that are not part of the main render but can help you adjust or tweak certain details by isolating a specific part of the image, like the "depth" pass which gives each pixel a different shade of grey depending on how far away they are from the camera (top left from the utilities quadrant in the image below), which you can use to add a foggy atmosphere to your scene, for example.

(Image is from the CG compositing series on compositingmentor.com)

Now, did I do all of that for the animatic? Fuck no, I'm not a compositor and by that point I had six weeks left to do the dang thing. However, I had learned Nuke in animation school, and Blender's compositor was quite similar. So to get that grey-blue/pink palette, here's what I did.

I started by creating different materials, each corresponding to a single shade of grey ranging from black to white (except for the one with the image textures of the posters and framed photos), which I applied to the models. Once that was done, I used the compositor to separate the render into two passes: The flat colour of each object not affected by light or shadow, and the direct light showing in pure black and white what was illuminated and what was not.

Then, I took the flat colour pass on the left, divided it into two versions, one for the light, one for the shadow, and applied a filter to recolour them appropriately.

By the way, that filter is called a colour ramp. It works by taking an image and using each pixel's brightness to assign them a position on a gradient defined by the user, which then becomes its new colour. A dark pixel will edge more towards the left of the gradient, a white pixel will be on the far right (appropriately), etc. Here's what that looks like with my two images and a more extreme example.

Once I had those, I just needed to merge them back together using the direct light pass from above as a mask, basically telling Blender "in the white area, I want you to show the pixels of the pink image and in the black area, the pixels of the blue one". And voilà ! I could now easily fine tune the colours by tweaking the gradients of the colour ramps. Then it was mostly just aesthetic polish, adding ambient occlusion, a vignette and chromatic aberration.

And now I had a set, I knew how to model assets in Blender, and I knew how to make them look good as background. And so it was time to figure out how to animate the characters.

5- A little bit of elbow grease... pencil

I hesitated quite a bit on whether or not I should draw them separately on another software or use the 2D animating tool native to Blender, IE the grease pencil. I went with the latter because it would make it much easier for me to do moving camera shots since they'd be actual objects in the scene, so no need to track them into the footage. Plus, that meant I could attach them to other moving objects if, for example, I needed one of them to ride a Vespa.

By the way, since the drawings have to exist as 3D objects in a 3D environment, they're not pixel-based but more akin to vector graphics, like what you'd get out of Illustrator. The strokes are made of individual points, each with an assigned radius defining the thickness of the stroke around it, connected together to create what looks like hand-drawn line-art with tapering. I just think that's interesting and good to know if you wanted to try it out.

I also spent a day or two making a rigged model to use as reference, because that's what I did for my previous project and it saved me a lot of time then.

However... in retrospect I probably shouldn't have bothered, or at least not spent as much effort on it. I knew how to rig in Maya, but Blender works quite differently and the rigify tool wasn't helping as much as I'd hoped. Plus, making a cartoony lizard dude is one thing, making a semi-realistic human with correct proportions, anatomy, joint placement and skinning is another. So even with the excellent mannequin rigs by Pierrick Picaut as reference, the result was... oof.

Having a human-sized and vaguely human-shaped model was still useful for shot composition but a simple sphere set at the correct height for the head could do the job equally well, and that's what I was using by the end. Nonetheregardless, it was a learning experience and now I kinda know how to rig in Blender, which brings me to another very important piece of advice.

Until now, this post might have given you the impression that this was all a straightforward process, that I just made a to-do list and followed it religiously until it was done. That couldn't be further from the truth. A good portion of it was achieved by scrambling, improvising, finding overcomplicated solutions for simple problems, using the wrong tools for the job, going back and forth between different steps of production, walking into dead ends and retro-pedalling... even when I did know what I was doing, the cost of that knowledge had always been paid during a prior project with equal amounts of trial-and-error (and lessons from my animation school teachers, you guys are the best). Fucking up is the most integral part of learning and creating Art.

But I'm sure you've heard it before. Don't worry, I'm not telling you "just be okay with making mistakes and bad Art" like this is something any of us has control over. This is more a reminder that all of the artists you love also made their masterpieces by fucking up repeatedly, both before and during their creation. It's hard to keep that in mind because we generally only see the final product and can only assume that every aspect of it is an intentional choice, translated perfectly from their mind to the canvas, page or sheet music. It was not. And you can bet your butt that what was lost in translation also keeps them up at night.

But with that, the ingredients were all in place. I made a few other sets and props that would be re-used in different shots, like the office corridor and the Vespa, and then moved on to the final stage of pre-production.

6- Buried in the workflow of time, in thy great name I pledge myself to darkness

My final task before attacking the main course was to set up Blender in a way that would minimise the amount of time and clicks between creating a new scene and beginning to draw. For example, there's a function that lets you save the current one to use as a default starting point whenever you create a new file, including the objects, the UI layout, the settings, etc. So I spent some time setting up a scene with a pre-made camera, several grease pencil objects for the characters and set-dressing, empty collections (basically folders that can contain other objects) in which to import the sets and props, tabs with custom workspaces adapted to each step of the process...

However, most of the refining and polish of that scene was done after I started animating, leading me to advice number whatever: Taking time to prepare and set up the terrain for yourself before you begin working on the main task is good, but it's possible to get caught in a "I can't start before I'm absolutely ready and everything is perfetto!" mindset, which is counter-productive and pointless. Because you can't figure what's the best way to prepare for the thing without actually doing the thing. So once you're 80% of the way there, just go for it, animate a shot, experience friction, take notes and go back to tweaking your pipeline for the next ones. Repeat until forever, because you're never done tweaking and that's okay.

Also, I took two days to learn how to use the asset browser, which lets you drag objects and collections from other Blender files. This might have not been worth it, but again, it was crucial in helping me understand how Blender works, especially when it comes to importing and linking data, which is very different from Maya. But thanks to that I knew, and that was a huge help.

Anyway apparently Mr. Eclipse made 5000 years' worth of solar eclipse graphics, which you can find the simpler versions of here