text was added manually. if it was all one pass there would be space for the text in the image. image models are sticklers for consistency.