gnarlyman at May 20th, 2012 14:52 — #1
Something I've always wanted to know, and now actually employ myself...
I want to download some mo-cap data files (actually, from some university's website that someone suggested on these forums in answer to a Q I had about free/low cost mesh and animation resources). What's the process by which one makes actual models and animations from these mo-cap data stuffs? Even better is if I could somehow intake the mo-cap data and produce meshes and subsequent animations procedurally.
rouncer at May 20th, 2012 16:01 — #2
i havent actually done it before, but i think its in a dynamic bone tree in the .3b file or whatever format it is, and you rig your model (link vertices to bones) before you can apply the animation to your model. you only get the animations provided in the animation file, if you want more keyframes, you have to manually make them in your own editor.
stainless at May 21st, 2012 04:45 — #3
Skinning and or rigging.
Basically you create a mesh with it's arms and legs spread out.
Create a skeleton that matches the one used in the mocap data and match it's position to the mesh.
Generate bone weights for each vertex.
The actual details of how to do each stage vary with your tool chain, it's very different in 3ds max from how you do it in blender, but it general, that's how it works.
Then when you animate the mesh, you position the skeleton, then use the bone weights to position each vertex relative to the bones.
In principle, it's really easy. In practice, a nightmare. I'm not a graphic artist and rigging a character usually ends up with me throwing things at the wall. The usual things, mice, monitors, people... occasionally desks ..
gnarlyman at May 22nd, 2012 21:37 — #4
So...does that mean it's not a relatively simple(r) process of creating a model with bones, and then loading the mo-cap data, having the software just produce the animation? Do you literally have to move the skeleton yourself for each keyframe of the mo-cap, or does the software (like Blender) do all of that for you? Little confused here.
And yes, I'm trying to avoid throwing stuff at walls. All of those things, lol.
Btw, is this why online deals like Mixamo.com (or whatever) are good stuff? What do products like that offer compared to simply getting some raw mo-cap data such as the stuff I mentioned above from that university?
reedbeta at May 23rd, 2012 14:40 — #5
I haven't had direct experience with the process of applying mocap data to an in-game rig, but from what I understand, it has to be pretty extensively processed and cleaned up to make it usable for in-game stuff. The process is growing more automated as time goes on, but still not at all simple. It wouldn't involve manually setting keyframes (hopefully), more about setting up the mapping from the mocap skeleton to your in-game skeleton - which might have different bone structure and different proportions. Mapping an animation from one skeleton to another is called "retargeting"; you can probably find some stuff on the Web about that. I don't know much about it, but from hearing other programmers discuss it around the office, it all sounds terribly complicated.
Raw data from a mocap rig is likely to be provided on a "here's what we have...good luck" basis, if you know what I mean. It'll be a lengthy, tricky process to munge it into something usable in-game.
gnarlyman at May 25th, 2012 23:58 — #6
Oh boy, yay. Lol, well, I guess I probably won't explore as much along those routes. It's true; such much more is becoming automated these days, and there are also so many libraries of pre-existing content already out there. I'm thinking of simply using such libraries versus creating such content from scratch.