{"id":1636,"date":"2026-04-07T17:44:21","date_gmt":"2026-04-07T15:44:21","guid":{"rendered":"https:\/\/stuff.tamius.net\/sacred-texts\/?p=1636"},"modified":"2026-05-01T21:32:53","modified_gmt":"2026-05-01T19:32:53","slug":"the-charr-cosplay-designing-the-frame","status":"publish","type":"post","link":"https:\/\/stuff.tamius.net\/sacred-texts\/2026\/04\/07\/the-charr-cosplay-designing-the-frame\/","title":{"rendered":"The Charr Cosplay: designing the frame"},"content":{"rendered":"\n
\n

Ta zapis je na voljo tudi v sloven\u0161\u010dini <\/a>\u2014 klikni tu<\/a>.<\/p>\n<\/blockquote>\n\n\n\n

Holy hell, this project has been “on hold” for a ridiculous amount of time. But it’s not like anyone noticed, so it’s fine.<\/p>\n\n\n\n

Previously<\/a>, I’ve been thinking about how big the costume should be, and ended on a note about anatomical compatibilities. I’ve also been using this image for planning:<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Notice any problems?<\/p>\n\n\n\n

That’s right. I’m comparing against the in-game human models. I should instead be comparing against a photo of me, and that changes a few things. Most notably, this “size comparison” also ignores that charr are about twice as wide as humans. This is a problem, because if the costume would be wider than me, then my arms would fit into the costume either very poorly or not at all. <\/p>\n\n\n\n

This means that we’ll need to change the aspect ratio of the costume as well. <\/p>\n\n\n\n

The problem is that we’re now 6 months late, so I kinda need to start figuring this out pronto. <\/em><\/p>\n\n\n\n

I started with praying to saint Billy Mitchell, and in a dream he revealed to me a shortcut to my goals. The idea of the shortcut roughly boiled down to: <\/p>\n\n\n\n

    \n
  1. Do the thing \/r\/GuildWars2 knows me for and extract the model from the game<\/li>\n\n\n\n
  2. Create a 3D scan of me<\/li>\n\n\n\n
  3. Overlay 3D model of charr on top of a 3D model of me in blender<\/li>\n\n\n\n
  4. lattice modifier<\/li>\n\n\n\n
  5. throw both into plasticity and sketch costume wireframe (and clamps that will attach the costume to my body) on top of imported 3D models<\/li>\n\n\n\n
  6. gg ezzzzz<\/li>\n<\/ol>\n\n\n\n

    It seemed like a good idea. Foolproof, even. <\/p>\n\n\n\n

    Getting a 3D model out of the game was very easy, especially now that I have the tools that automatically correct model distortions (I could also get the non-distorted model in T-pose, but for various reasons I decided to model the wireframe in the “default idle pose”). But that’s where #ez4ence ends.<\/p>\n\n\n\n

    STL of me<\/h2>\n\n\n\n

    The problem with my 3D scanner is that it’s designed for scanning of small objects from a distance of 10-30 cm, which makes it absolute dogshit for scanning me. <\/p>\n\n\n\n

    What’s my other options? <\/p>\n\n\n\n

    Well, photogrammetry is a thing. I could (and I did) ask my brother to take a shit ton of photos of me from a shit ton of different angles. Plan B was to record 4K video of a walk-around. Plan C was to have him do a 3D scan with his iPad. It allegedly has lidar.<\/p>\n\n\n\n

    On youtube, you have plenty of people saying how easy all three approaches are. <\/p>\n\n\n\n

    In practice: at work, I spend 8 hours a day sitting behind a computer. In my free time, I spend the rest of the day doing hobbies that also involve sitting at a desk, while also maybe eating a little bit too much. Do I look like someone who can hold a T-pose for 3 minutes, completely still, without shakes?<\/p>\n\n\n\n

    Not to mention that I have a serious lack of clothes that fit me tight and aren’t black, which can cause … issues. I’ve tried to mitigate some of the issues by intentionally setting camera to overexpose. The logic was that if I overexpose, I’ll bump those blacks into greys, which will increase my chances of getting something<\/em>.<\/p>\n\n\n\n

    \"0001\"
    Example image from the 4k 360 video.<\/figcaption><\/figure>\n\n\n\n

    Meshroom \u2014 the program that takes a bunch of photos and spits out 3D models \u2014 still pulled off a valiant effort … but the model is not of the desired quality. My arms are flat out missing, because as it turns out: holding T-pose completely still for 2-3 minutes <\/em>is incredibly hard.<\/p>\n\n\n\n

    \"slika\"
    The front side is actually relatively fine. However, the rear leaves much to be desired.<\/figcaption><\/figure>\n\n\n\n

    That was the result with a bunch of photos.<\/p>\n\n\n\n

    That’s what I got from the video: <\/p>\n\n\n\n

    \"slika\"
    Turns out that even at 4K, there’s not enough detail to get a good 3D model. That, or I shaked and swayed a lil bit too much.<\/figcaption><\/figure>\n\n\n\n

    Admittedly, the legs look pretty decent given that I was wearing black pants when taking photos\/videos. It’s certainly a lot better than the iPad scan.<\/p>\n\n\n\n

    Allegedly, the iPad has lidar n stuff. However, allegedly, lidar doesn’t work in object scan mode (or so my brother tells me), so it also gets to re-construct a 3D model of me from captured images. <\/p>\n\n\n\n

    Unfortunately, the iPad wasn’t set to overexpose (I’m not even sure if the app allows you to do that), which means on the iPad scan, my pants were vantablack black, kinda like a black hole that sucked in all the light. This resulted in a 3D model of me that was comically inaccurate \u2014 it made my belly twice as big as it is, and it made my legs into toothpicks. <\/p>\n\n\n\n

    There was a second scan attempt, with me sticking a bunch of masking tape all over my clothes. My brother scanned me, but had to leave before the processing finished. I’m still waiting on that model.<\/p>\n\n\n\n

    Will it .blend?<\/h2>\n\n\n\n

    While I’m not too keen on modelling on the current model of me, it’s still gonna be decent <\/em>enough for me to scale the charr to something a little bit closer to my slze. After slapping a lattice on the charr and dragging some points around, I manage to get something that manages to contain my body relatively decently.<\/p>\n\n\n\n

    \"charr-deformed-blender\"
    Seems like the overlap is relatively nice.<\/figcaption><\/figure>\n\n\n\n

    But at what cost? Well, the lattice is a giant, hard-to-manage mess:<\/p>\n\n\n\n

    \"slika\"
    Mom spaghetti.<\/figcaption><\/figure>\n\n\n\n

    But the end result? It seems fine<\/em>.<\/p>\n\n\n\n

    \"slika\"
    Game-accurate vs. body-compatible.<\/figcaption><\/figure>\n\n\n\n

    Sure, we did lose a lot of that hunch, but sometimes you have to make some sacrifices. Granted \u2014 if I were at a healthy weight, I could restore some of that hunch by pulling the belly back on the costume … probably. But it is what it is.<\/p>\n\n\n\n

    At this point, the fact that the 3D scan of me is missing my arms is starting to hurt a bit, so I tried to scan just my arm with my 3D scanner. How did it go?<\/p>\n\n\n\n

    Yeah, it made me remember how much of a piece of shit that scanner is. Scanner says it’s too close to the item. After moving it 5 mm, it’s too far, then the object I’m scanning is “out of sight” EVEN THOUGH I CAN FUCKING SEE IT THROUGH THE PREVIEW WINDOW<\/strong>, then the USB connector decides to be a fragile little snowflake and whoops, lost connection to the scanner. Upon restarting the scan, the scanner captures ONE (1) random frame of points in space that’s more or less a hallucination1<\/a><\/sup>3D scanner and the software to control it do not utilize AI, btw, so it’s not a hallucination in the AI sense of the word<\/span> \u2014 essentially, a garbage capture \u2014 and starts pestering you to “return to a scanned area”, even though you can’t because the “scanned area” is essentially a few random points. So you have to restart the scan a few times … and then it’s back to the “too close, too far, but never just right” game.<\/p>\n\n\n\n

    Eventually and through a lot of effort, I do manage to get a scan of my arm, aaand …
    <\/p>\n\n\n\n

    \"slika\"
    I marked my elbow by extruding some geometry.<\/figcaption><\/figure>\n\n\n\n

    Yeah. I need to correct for elbow position as well.
    <\/p>\n\n\n\n

    The Wireframe<\/h2>\n\n\n\n

    Now that we have a 3D model, we can go into a proper-ish CAD program and sketch a wireframe over the mesh we got from Blender. <\/p>\n\n\n\n

    With Plasticity, this is very easy: just import the mesh, and then start drawing curves on top of it. Thankfully, Plasticity will automatically snap the control points of the curve to the faces of your mesh, and it will never ever snap to some random point halfway up your mum’s fan\u2588\u2588y that you didn’t ask for for no reason. Snapping always makes sense and never ever makes mind-bogglingly nonsensical snapping decisions that make you selectively quote a certain Mentour Pilot vide<\/a>\u2014<\/p>\n\n\n\n

    \"kurwa\"
    It looks like a curve when you click it around the leg. It turns into a kurwa when you change your perspective, because Plasticity decided to snap one of the curve points somewhere random for no apparent reason.<\/figcaption><\/figure>\n\n\n\n

    Yeah, snapping in Plasticity is sometimes absolute dogshit.<\/em><\/strong><\/p>\n\n\n\n

    But before we do too much work, let’s ask ourselves a question. Few questions, actually.<\/p>\n\n\n\n