Hi Cory. Thank you for your interesting questions. They really made me think.

In my article, I mostly wanted to highlight how some UX tools, like a persona or a scenario, can help with composing an AI image.

Having clarity of what we want to AI generate before we start with writing the prompt helps to keep the renders on point, ie, we're not experimenting with AI, but looking for a very specific output.

Personally, if I'm composing a rather complex scene with a leading figure in it, I'd rather compose it in pieces. I feel that that gives me more flexibility with arranging the image. If we use a prompt only to render a complex scene, then we're relying on whatever output the engine generates, and might have to do a lot of editing out.

I'm sure that there are ways to craft a prompt to get pretty close to what we're looking for. Like in Firefly, we can input what we don't want in the image. Midjourney seems to output some amazing scenes.

There is certainly room for a lot of experimentation.