Part 6: When the First Good Image Isn't the Final Image

I liked the first “Lights Out West” cover. That could easily have been the end of the session. Instead, I used it as the starting point for a more specific question: could this feel more like the hand-drawn or painted album art associated with classic rock?

Boston's debut album was a useful reference, not because I wanted to copy its spaceship or its space setting, but because its image rewards a second look. The spaceship is built around the visual idea of a guitar. I wanted that kind of double reading: an image that works first as landscape and atmosphere, then suggests something else.

Reference the idea behind the reference

That distinction matters with AI. “Make it like this album cover” is a weak creative instruction. Explaining what you admire about the reference — hand-painted character, visual transformation, an object hiding inside another image — gives the AI a principle it can reinterpret.

The revised artwork landed much closer to the identity I had in mind. Then came one last small decision: remove “THE” from the top. The change didn't require a new creative concept. It was simply editing. Midnight Line looked stronger on its own.

The lesson

A successful AI result can still be a prototype. When something works, ask why it works and what still feels generic. And when you use a creative reference, borrow the principle behind it rather than asking for a copy.

Try it yourself

Find a piece of creative work you admire and don't give the AI only its name. Describe three specific things that make it effective. Then ask the AI to apply those principles to your own subject without reproducing the original.

For now, Midnight Line has a song, a frontman, a name, an emerging history and the beginnings of a visual identity. The next part will be written when the band takes its next real step.