Live2D works on what is essentially a blendshape rig system which is entirely dictated by mesh deformation and sprite layering.
It's all very simple stuff where the arragement of pieces is the entire product but it's not something I expect most people who aren't intimately familiar with motion graphics workflows to fully comprehend instantly.
I've had over a decade of after effects expereince so this is just natural to me but the idea of rigging 2D drawings isn't a common idea, especially with the L2D system which is essentially brute forcing the whole thing.
Eyelids are normally separate blotches of skintone which go between the eyebrow and eyeball layers and they use that and eyelash deformation to manually hide the eyes behind the skin before turning the opacity to zero and swapping in a closed eye sprite in place. There's a lot of simple but unintuitive stuff like that in the layering process such as how hair is arranged and how clothing with layers that go from front to back can be made, if an artist isn't familiar with the workflow nad just draws a PNG with no layering (as many do) then it's up to the rigger to separate all the parts and redraw what's missing from under the initial drawing.
That's fucked up but it's on the part of the person who commissioned the artist who didn't give them a detailed guide about how to separate their layers.
Once the parts are separated the L2D rigging part plays out the same every time. Map the sprites, map the blendshapes, bada bing bada boom maybe do a few line replacements and adjustments but it's all the same stuff after the parts are separated.
If the artist actually knows how to arrange a L2D spritesheet with proper layers it becomes piss easy for the rigger because they can just import a PSD file with all the layers and placements intact and all they do is rig the drawing with no extra work like filling in blank parts where they had to crop parts out.
For the sake of ease of example I'll use the V0.5 of my own model to demonstrate how much extra needs to be drawn in so that the blending of the shapes works correctly.
>https://files.catbox.moe/p93s13.png (embed) This is what the rig looks like with all the parts overlapping right, there's no distinguishing lines separateing the nose from the hair strands to the eyes etc etc.
>https://files.catbox.moe/rmb4vq.png (embed) This is how much extra gets drawn in behind, and that's for a super simple beta model.
If the artist doesn't do this the rigger has to, and if the rigger just knows how to rig and now how to draw you end up with major complications.
Hair gets cropped, clothes deform incorrectly, movements get significantly limited due to the sheer amount of missing assets, etc etc. The issues fall onto a matter of workflow efficiency.
Again, if the artist understands all this stuff then the rigger has a much easier time, if the artist isn't given explicit details the rigger gets fucked over. If the rigger is give a separated spritesheet with no arranged layers that can also fuck them over, if the rigger is given a good layered model and they make a shit rig, they're just a shit rigger which is more common than I'd like because most riggers don't actually know how to draw that well.
Part of the Live2D rigging experience is making expressions and angles for the model that weren't in the initial base, and you need at least a certain amount of artistic sense to get it right.
That's why riggers who do their own art are usually the best, they know what the drawing is supposed to look like at multiple angles. These L2Ds are a collaborative effort in the worst way possible, an artist may get incomplete instructions, give them to a rigger with subpar art skills, who then has to rig a drawing to be artistically cohesive while not actually knowing what the drawing is supposed to look like outside of the front view.
The responsibility is shared but the artists generally just do as their told, and when they're not told well they can't really be blamed. The riggers know exactly what their job is but it depends on the artist to make it easy.
>>6415658# Individual parts vary wildly from model to model, some will have as few as maybe a dozen moving parts and some get into the hundreds. >>6415705# [Streamable] 2021-02-16 00-03-05-1 (embed) Like I said this is V0.5 made in the uber limited free Cubism4.0, but it has the fundamentals.
>>6416405# There really isn't much to explain, without getting into the nitty gritty of L2D's rigging system with stuff like line tapering and the different deformation methods (none of which factor into the intial artist to rigger translation) it's actually a very simple system. Ive' you've ever used AE's puppet tool that's literally 50% of L2D rigging right there.I tried to make it a bit more readable
