A lot of people seem to think that generating AI art is easy.  You just type in what you want to see, click a button, and *poof* it's there, right?  Not really.  Not at all, really.  Getting set up is not too difficult.  You install some software, get things configured, find a good AI model, (or just use the SD models) and away you go.  You type in your first prompt and click the button and what do you see?  Most likely a bunch of humbled jumbled nonsense.  Unless you grabbed your prompt from some site like prompthero or lexica then you might see results similar to what that prompt is known to generate.  Or maybe you get something that's not even close and you think "what the fuck is this nonsense?"

There are a lot of misconceptions about AI art.  It's not easy or low effort.  I might make it look that way with all the content I generate, but I guarantee you that is not the case.  For each image you see, there were probably at least 5 others that didn't make the cut.  Those others either get thrown in the trash or, if they're good enough, then they get saved to go into my Nightmare Fuel or Derp galleries.  Not a lot of AI content creators post the misses.  It's great to see some of the other generators out there are posting some galleries of their misses as well.  

Crafting an ideal prompt takes studying, testing, and time.  And even then, SD just takes your prompt as a suggestion.  It still does whatever the fuck it wants.  As a prime example, right now I'm working on a gallery where I want wavy white hair.  Out of several images generated almost all of them have white hair.  But for some reason I changed one little thing and they started coming out with dark brown hair.  I didn't remove anything about the hair, I didn't add any other keywords anywhere that woudl generate dark colors, SD just decided that it wanted to create dark hair.  Often times what works great in one scenario does horribly in another.  Or, what works good with one model, produces absolute shit in another.   

Mangled hands and other fine details are a reality that we all deal with when producing AI images.  AI just does not draw hands well.  There are things you can do to combat it, but for the most part you just kinda gotta roll with it and hope you get one with a good looking hand.  Or, you can do what I do and focus on images where the hands are hidden or they are just not a focal point.  Even with that though, there is also other mangled anatomy that will happen, and happen often. 

So that's that.  AI Art is a lot of trial and error.  You throw something at the wall and see what sticks, tweak things and test again.  A lot of times my galleries start off with images that are 'test' images.  They're just the 'good ones' that I saw when testing.  I'm not going to do that anymore on themed galleries.  From now on if I have test images that come out decent I'm going to save those and post them in a Random gallery.  Then once I've tweaked the prompt and figured out the style I'm going for I'll start generating the gallery for real.  There might be one that slips through the cracks here and there, but it'll be okay.   

Hope this helps some of you who tried the AI thing and saw horrible results understand why, and helps the others understand that even though I produce a shit load of content, a lot of time goes into each gallery.  It's not all sunshine and rainbows.