With Ai as long as you are good at promps you dont need much skill
That is what produces slop, just putting in a prompt and using the result. The true art is getting the picture out of the AI that is in your head, or very similar to it. It's what drives you mad with AI; it's what wastes hours to get just one or two images, and the expense that goes with it. Because it seems to the artist that the AI is doing everything it can to pervert your vision into something unrecognisable, like a genie with your wishes.
Now the Lady-J picture, that was fast and required little skill or effort. I wanted Lady-J on a black motorcycle. Background wasn't important so it wasn't mentioned. No particular vision except a Terminator-style picture. Make sure it looked like Lady-J. This limited info was all I mentioned in the prompt, and this was the second or third image produced that was good enough. I usually have to do many more.
Often I have to keep it simple at first, then add in a feature or two, then change the location, then have the character doing something, like she is talking, so her mouth must be mid-word. Then turn her body slightly to one side or the other, then add the necklace that was missing, or get rid of an item she was holding that wasn't needed in that shot (a wand, for instance). Sometimes one picture takes over an hour and twenty to thirty attempts to get it right.
The Jane story used 12,000 credits on OpenArt over a couple of weeks. I actually ran out of tokens it used so many, and that was all images, no video. Hundreds of them. Even so, Willow is sometimes wearing her necklace, sometimes wearing a black choker, those are the things I missed. Many, many times her face changed so it wasn't Willow, close, but not her. She has a very distinct face, so when an image was complete I often had to redo it, asking just for her face to be swapped for one I uploaded, praying nothing else changed too. It's mind-numbingly hard.
So when people tell me things like it doesn't take much skill, or that it's just entering a prompt and it's automatically perfect — well, I know better.
Ten to one is a good ratio if you want your image to be anything like what's in your head. It always makes me mad when people say you don't need much skill, which is only true if you don't mind producing slop, or you're not willing to invest the time and money. Otherwise you need to be skilful at tricking the AI into producing what you need, or what you feel.
As you can tell, I've finished the week's work and have plenty of time to argue on the forum right now! 😀
