Darn I didn't see there was a second page to the thread until now 
You're correct when you say that a bitmap image is totally unsuitable.
Thinking it would be more comfortable to select folding edges rather than cutting edges, as they will _always_ be less numerous.
Pepakura looks like some fine software. A little worried about the output format though..
However, I am ultimately trying to get things done that are beyond simply this folding : selecting my object and applying a Doo-Sabin subdivision to it, I can observe the result is pretty close to what would happen when inflating the volume WITH ONE MAJOR DIFFERENCE : Doo-Sabin subdivision smoothes the vertices when inflating really makes those vertices stick out like the corners of a pillow.
Doo-Sabin could actually be a very good base to start from, because some of the work is already done (though I'm surprised the user has no control over the intensity of the effect). I'll post a picture, this might be clearer than I am likely to be with just words.
vertex_smoothing.pdf (Size: 14.85 KB / Downloads: 6)
For each vertex (or all together as a group), the red circle defines a "vertex recess radius", which in turn defines the center of the blue circles that will "bend" the vertex inwards, this effectively smoothing out the overall geometry without the cost of extra faces, requiring only that the material be bent a little. Using Bezier curves could be a pretty solution, but I am not too familiar with the mathematics of those.
Note the opposite is also true ; where there are no folds, a vertex can be pushed outwards, this exaggerating the pillow-effect. great for star-like shapes.
IMHO, that totally comes with the "need" to be able to inflate/deflate objects (also for volume calculation!!) in order to better adjust the radii involved. Of course, inflation and smoothing vertices out requires the use of non-flat surfaces, which could be problematic to render. But with a little bit of imagination, it's not hard to guess what the shape will be like after vertex smoothing and inflation ; I would say that as long as the vertices can be smoothed out in the resulting vector output that is totally sufficient.

You're correct when you say that a bitmap image is totally unsuitable.
Thinking it would be more comfortable to select folding edges rather than cutting edges, as they will _always_ be less numerous.
Pepakura looks like some fine software. A little worried about the output format though..
However, I am ultimately trying to get things done that are beyond simply this folding : selecting my object and applying a Doo-Sabin subdivision to it, I can observe the result is pretty close to what would happen when inflating the volume WITH ONE MAJOR DIFFERENCE : Doo-Sabin subdivision smoothes the vertices when inflating really makes those vertices stick out like the corners of a pillow.
Doo-Sabin could actually be a very good base to start from, because some of the work is already done (though I'm surprised the user has no control over the intensity of the effect). I'll post a picture, this might be clearer than I am likely to be with just words.

For each vertex (or all together as a group), the red circle defines a "vertex recess radius", which in turn defines the center of the blue circles that will "bend" the vertex inwards, this effectively smoothing out the overall geometry without the cost of extra faces, requiring only that the material be bent a little. Using Bezier curves could be a pretty solution, but I am not too familiar with the mathematics of those.
Note the opposite is also true ; where there are no folds, a vertex can be pushed outwards, this exaggerating the pillow-effect. great for star-like shapes.
IMHO, that totally comes with the "need" to be able to inflate/deflate objects (also for volume calculation!!) in order to better adjust the radii involved. Of course, inflation and smoothing vertices out requires the use of non-flat surfaces, which could be problematic to render. But with a little bit of imagination, it's not hard to guess what the shape will be like after vertex smoothing and inflation ; I would say that as long as the vertices can be smoothed out in the resulting vector output that is totally sufficient.