Abstract: Method, system, and computer-readable media for modifying images are disclosed. Main and reference image are received. Location points of content within main image that is candidate for match with reference image are determined. Main image and location points are processed using image segmentation foundation model to generate image mask. Image mask is received. First, degree of similarity between image mask and reference image is determined. Second, presence of reference image in main image at location points is determined, based on result of first determining. Modified image is generated from main image. Modified image is generated by one of removing reference image from main image when reference image is present in main image at allowable location, relocating location of reference image in main image when reference image is present in main image at unallowable location, or adding reference image to main image when reference image is absent in main image.