Meshflow vs Retopology . . . flow?
polycounter lvl 19
Something that I've been wondering about is how much of a balance there should be between proper edge looping, and following the mesh geometry when retopologizing a model.
I'd guess for critical area's like the elbows and knees there should definitely be enough loops for it to deform properly, but at the same time, if you have details that will benefit from the additional geometry and lighting from that geometry, is it worth it?
Does having assymetrical geoemtry on a place like the torso make it difficult for animation and rigging then? And if you're given a really high tri limit, like 15000+, where then does all of your mesh detail go?
Any thoughts?
I'd guess for critical area's like the elbows and knees there should definitely be enough loops for it to deform properly, but at the same time, if you have details that will benefit from the additional geometry and lighting from that geometry, is it worth it?
Does having assymetrical geoemtry on a place like the torso make it difficult for animation and rigging then? And if you're given a really high tri limit, like 15000+, where then does all of your mesh detail go?
Any thoughts?
Replies
It's mostly a model by model basis for me. If we're discussing the one above then I would say there are too many tri points where there could be quads. Which at worst would make skin weighting a little difficult if you're going by loops which isn't too much of a pain but its nice when people are considerate.
The edge flow as it is with a lot of tri points, could make deformation a little wonky but it depends on the range of movement required. It would require some fine tuning and even then there might be a few areas that end up in game and don't hold up like they should. Some of those areas might be to help define shapes that are only in the material so it depends on the whole package too.
Having a symmetrical mesh to weight is nice, because you only need to weight one side then mirror the weights. You can use the same trick for asymmetrical meshes also there's just a little more clean up. For the first mesh posted above I would suggest making it 100% symmetrical and build any asymmetry into the materials.
I would also want to take a closer look at the poncho and how it attaches. In general I would probably try to discourage extra pieces that connect between the torso and the arms as early as possible. With some extra bones, and extra time it would probably come out ok but it better be important to the characters design.
I think as character artists become a bit more seasoned and start thinking outside of the base statue pose they model in they start thinking about movement and rigging. I think its important for all character artists to at least have a good understanding of rigging/weighting and think about the motion and the deformation while modeling. They don't have to be an outstanding animator to know how a mesh will deform.
For areas like hands/fingers it can be a big time saver to have loops that terminate sensibly and are symmetrical for L/R hands. I would probably gut someone like a fish if they handed me a triangulated mesh for hands.
I take it then something closer to a lower resolution of the high poly mesh would be more suitable?
In this case, the design itself is pretty asymmetrical in the torso, but as far as polygon flow goes, this would work better?
What about tri-count then? If you have a lot of extra polygons to use, do they just all go to the silhouette and joints, or is there a proper way to use them for subtle internal details to catch lighting better?
As an aside, for the first guy, I did the textured version a while ago. And everything except for the goggles, helmet, and boot buckles, are one piece. Without there being floating geometry in the area does the poncho work better from a technical standpoint?
http://spiralpowered.com/images/Aang/avatarPresentationSheet.jpg
They make sense because they're reinforcing the material in this case. Maybe a few tris to make quads and put some poles in places that will help deformation. For each added cut you also have to anticipate how it will deform. Not only does it need to help the silhouette but it also needs to help deformation behave as expected. If its a wrinkle, it not only needs to bulge and sag, but it needs the surounding geo to help it behave like that. Depending on the folds it could require extra rigging to make it behave like it should.
In a case where the geo needs to really help define the materials and silhouette then I would just eat it and not even whimper. If its a case of tossing it in because there are tris to spare, then I'd probably grumble.
When you mirror skin weights, it takes the position of one vert on one side and looks for another vert in roughly the same spot on the other and assigns it the same weight. You can dial up and down how far it looks, anything it can't figure out what to do with it guesses or leaves alone. Its not a lot of clean up work but it can be annoying to look at a character that could be symmetrical and know that you have to skin weight both sides because its off. Again not a deal breaker just a bit more of a headache for the next guy down the line.
The points that concern me about the poncho can be negated with a little rigging and possibly a few minor edits to the mesh so it behaves like it would be expected. Or with a limited range of motion it might not matter at all. If he's a pilot that has his hands locked to a stick or only walks with his arms gently swinging at his sides then its probably ok as is.
If he puts his arms out like doing push ups, or above his head or into a T pose. Do they want the poncho to hang more naturally with little stretching, or is it fine if the parts are weighted to the nearest bones with no extra rigging? Just weight them to chest and arms and the stretching is just a fact of life for this character?
If its to hang, then each flap should probably be separated higher by the button and it might need to be separated from the arms so they can slide around under it.
There are a few ways to handle the poncho if its to hang.
- Cloth Sim and record the motion of bones that are attached to the cloth sim. These bones then drive the final low poly mesh.
- Add a few bones for the poncho and point constrain them so they average the distance between two bones (chest/arms).
Probably provide a system that sits on top of the base rig, so the person can nudge them around manually if needed.
The best advice I could give would be to rig it up and see how it deforms. Now that the materials are set and baked its going to be a bit of a pain to rework the topology and mesh flow but it would be a great exercise in what to expect.
For me the workflow goes:
Low base (quick unwrap) > high sculpt > low retopo > rigging and deformation tests > geometry edits and final unwrap > final rig > bake > materials.
For the coat model, the requirement was to bring the entire model up to above 10000 triangles, so its kinda both that its helping with the shape, and that there will be tri's to spare. I guess if I build a lot of the bulge edge loops, and normal loops for matching up the silhouettes that should be fine.
For the poncho kid, I don't know if I want to go through rigging for him at this point. . . but he would make a good test bed for experimenting with rigging because of the things you pointed out. Especially since that's an area I'm weak in.