So I was thinking 3D, with a full controllable camera, but I think I could get away with more doing a scene like this:
I think if the camera were closer, there'd be higher expectations for animations, and that's the one thing that I'm terrified to touch.
Doing this way, I think I could probably do a lot with a little basically.
I think the hardest part will be setting up skinned meshes and animation systems on the programing side of things. If you're able to get that going, I think you'd be able to make serviceable animations in blender with relatively little effort.