The models all are created from paths, I have no experience with "proper" 3D modelling (Blender etc) so I wouldn't really know how to import new ones.
- Path0, 1 and 2 respectively are the silhouette of the car at the center, a bit outside the center, and at the left/right sides.
- PxLengthFactor is the downscaling factor from path coordinates to model coordinates (the paths are pretty big thanks to how the grid is defaulted).
- W0, W1, W2 and W3 are how far left/right from the center the paths are placed when used as a guide for the model. These are raw model coordinates, they don't use the PxLengthFactor. So for instance having a higher W0 means the cockpit is wider, having a W3 much higher than W2 makes the car have bigger wings, and so on.
- W3 must always be greater than W2, W2 greater than W1, and so on. (You don't NEED them to be greater, but the model will clip into itself if not).
- The texture is mapped like this: every horizontal 25% of the texture is used for the section between two W*s. So the first 25% of the texture are used for the section between -W0 and +W0, the next 25% are used for the section between W0 and W1 on both sides, and so on. It's stretched evenly from the front of the vehicle to the back (the top of the texture maps to the start of the path, the bottom to the end of the path)
- Currently the model doesn't affect the collision mask at all, the cars all use the car sprite as a mask. You could extract the car's length from the modelling script by storing the "xx" variable of the first and last iteration (first iteration's lower value + last iteration's higher value, it's an array) and the W3 (outermost width)... those will give you the bounds of the car's model's bounding box, and then if you used an ellipsoid collision you'd get an okayish result. Probably would be the easiest to make a circle-shaped sprite for collision masks (with Ellipse collision mask settings, sprites default to "rectangle") and then scale image_xscale / image_yscale based on the bounding box size so the sprite covers the same region (e.g. if W3 is 8 and the sprite is 32x32, image_yscale should be 0.5 because (8*2)/32 = 0.5)
Hope this helps :)