1- So, a "consistency" test, meaning renderings from different viewpoints with the same characters. It's not perfect, but the method is to create two 3D renderings (here in Rayscaper) and place the image in SketchUp. Then, I ask D5 Render Lite's AI to add characters to the first viewpoint. I also give D5 the same image as a reference. Then, I take the least bad result and create a Photoshop montage.
2- To get the same characters from a different viewpoint, I place the new rendering with a new viewpoint in SketchUp and give D5 Lite my latest image with the characters. I then request the same characters in a different location.