I'm writing a program that will load an animation from a file using assimp. According to this answer "The transform we use for skinning is (B_keyframe * offsetMatrix), where B_keyframe is the global transformation of a bone at some target location, provided e.g. by an animation clip." so... offsetMatrix is the inverse of the global transformation of a bone which would be computed by multiplying by the mTransform of the parent all the way up to the root aiNode.
But what use is mTransform then if the mesh is already in bind pose when I load it in?
For example, lets say a cube is parented to a sphere. The root aiNode has the sphere's transform, and the child aiNode has the cube's mTransform. If I were to edit this in blender and change say the scale of the sphere would that change the vertices in the aiMesh struct? Or just the sphere's mTransform? How would I render the cube in my program? Would I multiply the mTransform of the sphere with the cube's mTransform? Or has the scale already been applied to the vertices of the aiMesh belonging to the cube?