Our method can produce hybrid images, which change appearance depending on the distance (or size) at which they are viewed. We do this by conditioning low and high frequency components with different text prompts. These images were first proposed by Oliva et al.
