I've noticed a lot of art-generating AI popping up lately, ones that take basic descriptions of things, and try to generate an image from that based on thousands of other images it finds or has analyzed before.
And of course, it got me thinking of seeing if there are any you guys find to be better than others, and/or if anyone has had a good experience with getting one to make hypno pics in the past.

I did "Kaa hypnotizing Kim Possible" and got a picture of Kim slurping up Kaa like spaghetti.

The hunter became the hunted.

LMAO go figure. That's pretty funny

AIs aren't good at drawing characters yet. So far they're only good at drawing landscapes

I've been trying a bunch of hypno-horny prompts on a local Dall-E Mini and I've had interesting results but nothing really good. What I'm really waiting for is for Dall-E 2 to go public with fully trained models.

On another hand, I'm building Tensorflow datasets from a selection of Hub posts and other boorus too, so I'll try to train image-generating AIs from them.

According to many of the artists at my workplace, that is false, lol. With as many AI as there are out there now, they all have their skills and shortcomings, but some are in fact capable of drawing people. Which ones, I was not told, but I'll surely be prodding for the info.

Looking forward to seeing some results from that! Good luck!

From what I've seen, a lot of AIs still have trouble generating realistic faces, but can be pretty convincing with more cartoony/stylized art. It would be interesting to train one on a sufficiently-large dataset of my own drawings & see what it comes up with. I imagine I'd still have to clean up the results quite a bit, but it could be interesting.

I haven't used it, but it seems like the program EbSynth is quite effective for generating animation. Essentially you just input the frames from some video, along with a couple of your drawings that sync up with certain keyframes (the drawings don't have to be outright tracings of the video frames, but the characters should at least be in roughly the same poses/positions), and then the AI generates new drawings to match every other frame. If you had a friend or lifedrawing model with roughly the same bodytype as your character (they don't need the same clothing or hair), you could do a photoshoot with them (letting them know what this would be for, of course), do a character drawing based on one or two of the photos, and then get the AI to autogenerate the rest.

Dall-E 2 is some actual psycho shit but don’t expect a viral-ready Web 3.0 AI application to allow you to make porn. i might be thinking of Google but i’m pretty sure the devs said they took extra care to train the ai on only the most heckin wholesome 100 funny animal pics and cut out all that mean nasty porn and anything else provocative.

I do imagine that within five years there will be some absolutely insane AI visual tools freely available. AI text generation can already be very high quality, assuming you know how to guide it properly.