Jump to content
  • entries
  • comments
  • views

BodySlide TODO



BodySlide is a slick piece of work - it reflects some incredible patience and dedication in its construction.


Nevertheless, there's always room for improvement. And, there's a few aspects of its UI which I would like to fix (if I would work up some of that necessary patience and dedication...):


  1. Multiple nifs (sometimes from multiple mods) can write to the same destination, conflicting with each other. BodySlide lets you make choices about this, using a popup modal dialog, when you are doing a batch build, but it's hidden information otherwise. Also, personally, I would really like to be able to examine the output paths (three of them: the *.tri, the *_0.nif and the *_1.nif).
  2. Some outfits have zaps, but most do not. These can only be used interactively and finding them among the hundreds (or thousands?) of outfits can be daunting. 
  3. There's some normal maps which you can select for some presets. These do not adapt when you change the sliders significantly.
  4. There's for example the Curvy and the Curvy (Outfit) presets, Skimpy clothing which includes a body mesh could use the Curvy (Outfit) preset but if you are using the Curvy preset for the body mesh then any skimpy outfits must use that same preset (unless its thick enough to hide the resulting clipping problems, but that does not work well).


Here's my current ideas for those:


  1. For the Multiple NIFs problem: I'd like the BS UI to show me the output file paths with a popup menu to jump to any alternate source outfit/body for those paths. 
  2. When selecting an outfit, the count of the zaps (and maybe the number of sliders) could be displayed in the drop down. And/or maybe the zap choices could be preserved for each model so that batch build could repeat those choices. (And having "zap based auto groups" would also be nice -- so I could see all outfits with a particular zap name.) 
  3. Conceptually, bodyslide could generate normal maps (for the body at the very least). Conceptually, it could surface subdivide the model a few times to smooth it, and then interpolate to fill in the rest. Would this be good enough? Maybe?
  4. Conceptually, BS could recognize skimpy outfits and either give you an auto-group mechanism to batch them separately or ... whatever. "Skimpy" could be automatically handled by running some collision tests but this should be overridable in outfit studio for exceptional cases.


And then there's some additional functionality which probably belongs elsewhere. 


Like: tight clothing tends to run into animation problems with clipping. 


In particular, skimpy tight clothing tends to "not work very well" because for "low poly" game clothing models to work with low poly body meshes, the vertices need to be aligned along a normal vector (and weighted properly, of course). But if there's a tool that already deals with that issue, I do not know about it. This would also apply to clothing which only shows a little skin and is only tight in a few places. (Usually that includes the necessary "skin" in the model, but ... those are "optimized meshes"... skip down for my hopes there.)


The concept is simple, though a bit tedious: if you subdivide the clothing surface sufficiently, you can intersect body normal vectors from each of the body mesh vertices into the clothing mesh and rebuild the clothing mesh using the resulting vertices. But this gets you into some problems: interior angles, cloth edges and optimized meshes. Both of these probably require some artistic intervention.


Cloth edges in particular might need some support for editing the texture (with transparent pixels where the body vertices do not align with the intended cloth edges). So that's tricky to think about...


Interior angles (like armpits) probably just need a little touchup with the usual blender tools. (And maybe blender has a way of doing this vertex rebuilding already? If so, it's not well advertised/documented, from my perspective... But maybe that's just me?)


And, optimized meshes? Often a clothing mesh will have some completely unrealistic section for hidden parts of the clothing -- if it actually existed, it would slice through the legs or something. Conceptually, this could be detected and automatically removed, or trimmed, or something. But the procedure here is different from the "tight clothing" procedure even though the mesh "touches the skin". And there's a tricky conflict here between "clipping" and "cutting". I wish I could distinguish these cases, and remove the "cutting" cloth.


My ideal here would be to create "skimpy" but "not very skimpy" clothing and armor. Typical skyrim (or fallout) skimpy armors are kind of ugly, in my opinion. I want something that shows some skin but not to that extent. I think the current trends are largely a consequence of the available tools, and I would like to fix that.


If I could muster enough patience and dedication... There's some algorithmic challenges here that I've glossed over.


For example, how do you match up the nearest vertex in the body mesh to a corresponding location in a cloth model? One way that works, of course, is to do an exhaustive search. But that's too slow for most of the vertices. But, once you have one vertex pair like this, you can use the corresponding uv coordinates to find adjacent locations, which should cut down on the search time by a lot.


For example, how do you find the relevant "closest point" in concave locations? Well... maybe we goof up there and rely on a human to fix the resulting mess?


Did I mention that this would need some work?

1 Comment

Recommended Comments

  • Create New...