combining time-based diffusion with an image generator powered by deep learning