wolgade wrote:Thanks for publishing the results of your research. Especially "*shirt" looks interesting. I guess you had your share of trial and error before you came up with this idea.
There was quite a bit of trial and error involved in making the shirt buttons stay round and on the shirt. It's interesting to see the results of different vertex group assignments.
Even when the results are a big ol' mess, it's still informative to see where the meshes break and warp on various body types in different poses.
I did all sorts of different vertex group combinations on the buttons and/or parts of the shirt.
All buttons in one *group keeps the buttons round, but they all float above or sink into the shirt in a nice rigid line -- looked pretty bad.
Assigning each individual button to its own *group and then assigning the three closest vertices on the tights helper for each works well on the average-sized female human, and also keeps the buttons round, but the buttons float and/or sink when the human's age, weight, proportions, etc, are changed.
I looked at how the buttons were sinking and floating between the breasts as I made the human heavier or thinner, adjusted the height and proportions, and played with the breast size, firmness and vertical position sliders, and decided that the buttons need parts of the shirt included in their vertex groups in order to stay stuck to and above the shirt.
At this point, I still had seven different *button groups -- one for each button -- on the shirt, and seven three-vertex *button groups on the upper tights. I started out with one shirt vertex for each button and the results were very much the same as using only the buttons.
So, I kept adding more and more shirt vertices behind each button, until I had seven *button groups that included each of the buttons as well as the all the placket vertices behind each one.
When I noticed that each button/placket *group seemed to break into seven separate rigid squares, I thought that maybe I ought to treat the buttons and placket as one thing instead of seven separate things.
Then, I selected all the buttons and the whole placket, assigned all them to one group called *shirt, assigned three mid point vertices to the *shirt group on the tights helper, and ran it through MakeClothes.
I was surprised that it worked so well, and on so many different body types.
The jacket buttons worked better separately because one is above and the other below the waist -- where the tights and skirt helpers meet.
For the brooch, I just selected the whole brooch and then the three vertices behind it, and that worked just fine on most humans.
In some instances, depending on neck size/shape, the very top of the brooch sinks into the shirt collar a little, or floats a little too far away from it.
I tried the same trick as I did with the buttons, selecting a few vertices on the shirt behind them to include in the *cameo group, but that caused the shirt collar to deform in weird ways and parts of the brooch still disappeared into the neck along with the bit of the collar assigned to the *cameo group.
So, sometimes it's a matter of trading off one set of flaws for another and choosing which set of flaws to go with seems to be a matter of which ones are easier to fix in Blender.