There can be more focus on the uploaded image and how AI generate the background. most places there are scope of improvement in blending both the images. there is also good area of improvement needed in terms of selecting some background (for eg: temple, i am getting the same background even after searching with multiple keywords)