Impressive, again. That's why I asked even after reading your comment. I'm sure it can not be easy. Most images I see generated with AI are not even close to what you achieve
Well, I have been on Midjourney AI graphic engine since it first became available a couple years ago. I have done some 15,000 separate images, often taking many attempts to correctly prompt an image I find acceptable. It has save me many dollars from having to licence images online. I blog to around 20,000 folks and I like to illustrate my essays with an appropriate image. Licensing brought sub-standard images, while prompting properly produces good images. These are illustrations and not attempt at fine art, but they help a lot.
I don't make a big deal of my AI graphics. These are just illustrations and I don't worry over them as I do photographs. It is just easier to get reasonably good at prompting useful illusttrations than it is to sort through hundreds of thousands of commerical images, looking for one that will more-or-less go with my blog or essay, and then have to pay for it.
By creating my own images, I get illustrations that are much closer to what I need than any licenseable images. My readers can't even tell that many are not actual photographs although on each AI image that I post I clearly state:
[Midjourney graphic prompted by me.]
In a couple more years, we will not be able to tell the difference between AI illustrations of camera born images. Both are constructs. As they say, "Beauty is in the eye of the beholder." Also, I am kind of keeping up with what's happening in image creating.