This the second vtuber model that I've created which uses traditional animation techniques for the facial features, rather than warpers and blend shapes. I started this model in 2024, but added more mouth shapes this year to give her a broader range of expression.
She's twitchy and glitchy, which is due in no small part to my crappy face-tracking setup, but also for reasons I'll explain below. Luckily, for this particular design, it works.
Doing models this way is very time consuming; doing the drawings is the easy/fun part, but the rigging process is a real slog [who am I kidding? The rigging process is always a slog, no matter how the assets are created!]. And, unfortunately, the results aren't perfect under any circumstances because the vtuber software that exists right now is not designed for traditional animation.
I've rigged Mugi (and one other model) with Live2D Cubism as well as Inochi2D, but both produce the same undesirable artifact: cross-dissolved textures whenever an incoming tracking parameter value falls between the breakpoint values established in the rigging software. The result is blurred/flashed areas of the face, and it's not a good look.
I suspect that this could be resolved not with the rigging software, but with the performance software if it allowed you to constrain the precision of tracking data.
Facial depth is established with parallax, which puts a significant constraint on how much pitch/yaw you can do with the face, but I think the effect looks pretty good. Here's an exploded view of the model:
And here's the original sketch for the model. A bit more intense than she ended-up being once completed. Although... is she complete? I could keep adding frames to this model indefinitely. There's just something weirdly captivating about her, don't you agree?
©2025 LÆMEUR. Most rights reserved.