Some segmentation stuff seems close but they only separate objects, they don’t add objects. Best workflow I can think of with minimal editing is doing I2I, creating a mask for the new object, then using the mask on the original image.