There are probably many simple yet effective applications of diffusion models such as this. Many SaaS businesses and startups waiting to happen.
Randomly stumbled over this video. Apparently uses the Midjourney[1] bot to generate the images used to make the music video from the lyrics of the song.
On the one hand the generated images can appear to capture some abstract ideas, in other cases it's terribly literal. Still, a fun experiment I thought.