The lace (texture) came out exceptionally well.
Did you shrinkwrap the dress or pinned and let gravity do the rest?
The lace and silk are all procedural and because they're simple forms I didn't even bother to unwrap them - it looks good enough with simple mapping.
As for the dress I prefer much more to scale the figure or morph the pose inside of the dress rather than scale the cloth itself. Scaling the cloth introduces additional unwanted stretching forces in most cases, so scaling the figure seems to be more natural. I only needed to be sure that the figure lowres proxy doesn't intersect the cloth in any location nor the cloth intersects itself. It wasn't that simple considering the cloth is built with many layers (but still a single mesh!) and the figure wasn't in the neutral pose. If I used the neutral T-pose at the beginning and morph to final pose after the cloth's already been settled I probably wouldn't have a problem with the strap slipping from her right arm. So there was a scaling of the character over the span of 40 frames and scaling of velocity from 0 value (not falling due to gravity at all) to 1.0 value between 20 and 40 frame.
There's one interesting thing to be noted. I've been building the mesh wanting it to have many layers but couldn't handle higher mesh resolution. When I finished the lowres model I wanted to subdivide it in Anim8or... and it started to fall apart and the spaces appeared near the sewing edges. I expected the problems to appear because I knew I've built non-manifold mesh at the edge where the lace is joined with the main dress (think of two-headed snake).
I wanted to perform the subdividing in MeshLab but it didn't allow me to do so complaining about non-manifold mesh. At that moment I thought that my work was doomed and I'd have to redo everything at higher mesh resolution by hand
Hopefully the subdivision in Blender did work as expected and saved the project from being terminated in worst case
Anyway - I don't know how to pin the cloth element edges while still not being rigid differently from making the sewing edges non-manifold. At least now I know it still works correctly during the simulation.
