I had a phone conversation couple of days ago with a not-very-internet-savvy friend and somehow we ended up joking about DJT and Kristi Noem mud-wrestling, or something like that, and I informed him that AI could generate it. This was based on the cool stuff I have seen on this site.
Anyway, he requested a video with this prompt: “DJT and Kristi Noem mud wrestling while Ted Cruz gives commentary.”
I thought I’d quickly be able to send him something to make him laugh, but apparently I’m not so internet-savvy either. Anything I have found has either produced horrible bad results that don’t even look like the persons in question, or I’ve gotten a message that the content I’ve suggested in unacceptable.
Does anyone have a line on a better generator? Images will do if video is asking too much.
Just…don’t do it. I know your intent is benign, but this is just normalizing the entire notion of “deepfakes”.
Stranger
Point somewhat taken, but are you saying we shouldn’t use AI technology period? What if I was looking for a way to make a picture of “three kittens eating ice cream on a UFO”? Does that bother you?
In any case, I would very much like links for the latter even if the former is questionable.
“Three kittens eating ice cream on a UFO” isn’t portraying actual people fraudulently engaging in some activity. I know that you intend this image as a kind of satire, and if a political cartoonist were producing this as a clearly fictional work of their own skill I would regard it as mildly amusing (if in somewhat poor taste, but then, what political satire isn’t?); however, I think that using generative AI to produce quasi-photorealistic images of actual people doing anything is essentially a deepfake even if the intent is not to deceive.
Just my opinion, and I know that many people here and elsewhere vehemently disagree to a point of responding with vitriol and outrage that any limitation on the use of generative AI is some kind of reactionary technophobia.
Stranger
As for your original question, I doubt any current models would create an image of what you want well. Even with systems that don’t block celebrity names and even with one that knew all three names, the more subjects in a scene the more they tend to be blended together. And complex poses tend to get Picassoian.
Thanks! That’s hysterical!
I’m pretty sure all those images are real and kittens are actual extraterrestrial beings who are here to consume our reserves of ice cream and mutilate our cattle. Also, why are they so into pastels?
Stranger
And some of those aliens are nekkid! 
Are some of the kittens polydactyl?
Unstable Diffusion has none of the usual censoring that some of the other image generators have. It will (try to) create images of anything you ask for. Depending on the subject material, they may or may not look tremendously believable.
