Finally tried out Stable Diffusion, here's the best of it. Still getting the hang of it, but some of these aren't bad if you don't look too close

These look wonderful how did you do it?

Thanks!! Here are some other new ones.

You can learn to do this too! This is all using Stable Diffusion alone, no retraining with new images or coding. I signed up for Dreamstudio Beta, which is free at first, and used the options there.

You never get a good result on the first try. Dreamstudio has options to erase parts of the image so you can do "inpainting" and fix them. I'm still learning how to do this well, but it's really important.

To me it feels more like a game than a magic wish-granting machine. You think about what you want and then see if you can get there, or you see a result that's different than you expected but you can refine and work it out. You have to improvise along the way. It takes a while, like I can "play" this for a couple hours and get a few images I like.

The main thing is in writing the prompt, and running and rerunning and refining what you get. There are a ton of tutorials, explanations, videos etc. on this out there. It's really important to learn about the "negative prompt" which tells it what *not* to draw.

Stable Diffusion won't do everything, and some wam things it does better than others. For one thing it really prefers to draw clean faces, which kind of cramps our style here ... so you have to be specific that you don't want that. And even then it doesn't always work. Of course it won't do nsfw images. Those of us who like our wam pg-13 are at an advantage

The basic steps I follow so far in Dreamstudio are:
-set the model to v2.1, the newest one isn't the default.
-write a first prompt, see some examples below. I put the basic settings at CFG scale 15, Steps 10, Number of Images 6.
-that won't give you great images, but you can see if you're on the right track.
-when you see one you like, pick it for the image editor.
-run 4 images on that, with "initial image mode" and image strength between 10 and 30, with 20 or 30 steps. Usually tweak the prompt too. Do this a couple times, see if you're getting warmer.
-when you're pretty close, switch the model to "Inpainting 2.0" and use the masking tool to erase the parts of the image that you don't like. (usually extra limbs, wrong hands, deformed faces, other nightmare fuel )
-run some images on that with 30 or 50 steps. Higher steps cost more credits, but I usually don't see better results above 50 or 60 honestly.

A couple of example prompts I've started with -- these won't "work" right out of the box but to give you an idea the level of detail:

luscious glossy HD digital portrait photograph, beautiful college girls having a messy food fight with desserts, fun and playful, embarrassed, creamy, gooey liquid, splattering action shot, girls with pretty faces and lips | gross, vomit, red, brown, blood, dry, glasses, yellow paint: -2.0

gorgeous slender woman with beautiful arms, playful and embarrassed, messy sloppy thick full mask made of shaving cream, shaving cream piled on top of hair hiding her hair, thick shaving cream hiding her face, luscious HD digital painting, liquid sloppy reflecting light, trending on artstudio | scary, horror, dark, commercial, fantasy, cartoon: -1

full body photograph of a playful young woman getting slimed like Nickelodeon, shiny liquid slime pouring down from above, glistening shiny liquid, big splash, splattering over head and face, HQ, sharp digital photography, gorgeous, beautiful face and lips | gil elvgren pinup drawing of a girl: 0.2 | blood, horror, dark, dull, water, green, clay, blobs, paint, cotton, fabric, fairy, fantasy: -1

This is really cool