So this is something I started doing recently. I feed all the lyrics from a song I like (points if its descriptive) into an AI image generator, see what it gives back, and turn that into a lyric music video. Immigrant Song – which, besides Battle of Evermore, is probably my most listened to Zeppelin song – is one of the first few that I worked on.
You can usually figure why a particular lyric returns a particular image, but the most interesting is when you give it vocalizations. Towards the end "Ooh-ooh, ooh-ooh, ooh-ooh" is repeated several times. I'm not sure how Midjourney figures out what image to generate based on these lyrics but so far, they return a lot of blue & red.