Well, that topless image certainly had me clutching my pearls, and nearly fainting!
No, I was shocked only in an unexpected sort of way, not morally offended. Nudity should be permitted in art, though I can understand filtering it out. It’s too easy to face-swap someone IRL into a pornographic scene without their consent and cause problems for them online.
Besides, I can easily put clothes on a rendered character with Adobe Firefly’s in-painting. Photoshop is especially prudish concerning nudity.
In a local Facebook group, someone posted photos of a flock of turkey vultures perched all over a water tower. It was news to them, but locals know the vultures have been living on that tower for years, and that attempts to chase them away failed.
In the comments someone said that it was unacceptable and somebody needed to call the police. I then tried generating confused police attempting to arrest a turkey vulture near a water tower. This is the image I used in my reply.
If that’s someone’s goal, they can pretty easily find a trillion nudie sex pictures online without needing an AI to generate some. They’ll even have the right number of toes. Unless you have really specific needs (“Let’s discredit [person] by saying we saw them nude, riding a green bear through a Safeway!”), there’s little need to bring Art AI into it.
In super random news, I got an email from NightCafe today congratulating me on five likes for an image that I probably made in March 2022. And it looked like an image made in March 2022, if ya know what I mean.
Also I wrestled with MJ too long for a specific image of a wizard without a beard. MJ has some firm ideas on what a classic fantasy wizard should look like and it includes a beard. Bing could maybe do it but with a worse looking image overall. [Edit: Actually, Bing also added beards and the images are just bad for what I requested]
Sure, you can find a trillion porn photos online, but not generally depicting specific non-consenting individuals. And, sure, there have been ways to identity-swap photos for a long time, but not as easily or accurately as it is with AI, where any chucklehead can do it with ease.
And, I can think of plenty of groups of chuckleheads who would stoop to tarnishing other folks reputation—disgruntled, vindictive spouses going through divorce, fighting for child custody, for one.
It’s unfortunate that artists have to suffer because of chuckleheads, but there you have it.
That’s why you’re face-swapping though. Let’s step away from nudes for the sake of site moderator stress levels and say I wanted to fake someone being a McDonald’s employee for whatever reason. And fake them well enough to pass an actual inspection, not just for lulz on Facebook. It’s way easier for me to find a photo of a real McD’s employee and use either an AI Faceswap program or manual Photoshop to paste them in than it is for me to wrestle with an AI art program to generate a realistic looking McD’s uniform and then swap in the face I wanted.
One of the few situations where going “Full AI” would be better might be when you have a program that can generate the image with the appropriate face from scratch. That’s something you can manage in local Stable Diffusion via Loras or Face Swap plug-ins but not something you can do convincingly with Midjourney. Another might be if you needed a very specific setting/scenario you’re unable to find by canvassing the internet for McD’s employee photos but, in that case, you’re still probably better off just Photoshopping something than trying to convince an AI to render that very specific scenario.
A balding and beardless man in his early 60s dressed in black and purple academic robes with yellow trim, clean-shaved and gray-haired, holding a black & silver walking stick and a magical tome, painting in style of Larry Elmore and Clyde Caldwell and Jeff Easley
Example of results that would be nice aside from the beard:
I originally had “wizard” and tried switching to just “man” thinking maybe it was “wizard” that evoked a beard each time but no such luck. Also, I could pull it into SD and mess with in-painting it but then it’s a different pain to try to emulate a style in AI Art Program #2 that matches the one from #1 and I’m just not that committed
Interestingly enough, the image I’m using now (it was for a Roll20 token/char avatar) was from Midjourney around v3 but v6 is more insistent on the beard. It’s not actually “important” (my current image is fine) but was me wondering what v6 would look like then going down the beard rabbit hole.
A while back Playground AI removed their Playground v1 AI option. I noticed just recently that they have added Playground v2. I’ve tried a few prompts. It is very different from Playground v1. A year ago it would have been very impressive. Now, it is mediocre. Not nearly as good as DE3 or late Midjourney, but not the worst thing out there either. You can do celebrities with it at least.
Here is “Photo of Mulder and Scully looking with concern and worry as they argue with a xenomorph. Side view”
And DE3 for comparison.
Here is Jennifer Aniston playing chess with Donald Trump.
After a few tries I developed a prompt that produces good hands with hands, but no amount of trying so far can produce hands with hands with hands, except one time I got one image with a hand hand hand on one thumb thumb.
I was revisiting this human generator. I didn’t realize the first time around that I could add text directly into the prompt rather than just using the pull down menus. It uh… also allows you to use the word ‘nude.’ Just leave all the clothing options clear.
Funny how someone that has never used AI might mistake an image that you get as containing details that you actually asked for.
These are all recent failed (to match the prompt, at least) Bing/DE3 results for “photo of a banana eating a banana, grainy disposable camera snapshot” with different years added to the end.
I decided to try to describe what I actually got for one of the images and plug that in and see if I could reasonably reproduce it. So here are the results for three different variations on “A tight closeup of just a hand holding a Polaroid instant-style photo. The photo shows the following: a hand holding a peeled banana by a table with shallow bowl of cereal on it and the Kremlin in the background.”
