I’m suprised you were able to get such a good upside-down cocooned image. It had great difficulty with the idea when I tried to reproduce that XKCD image.
At first it had her dangling from the moon. I had to move her to the palm tree.
I get things like a cocoon much too small for the person and a right-side-up head on an upside-down body.
Not the thing to wear to the first day at a new job, even if the dress code is casual:
Context: An attempt to illustrate a roleplay story scene. That slogan had been on the shirt she’d worn yesterday, which the AI had invented. In my prompt, I’d told the AI the character was picking something different to wear today to the new job, without being specific enough, I guess.
I put a joke over in the More Jokes thread.
Going forward I told it hey, that’s not something she would have picked for her first day and her appearance doesn’t match well with the earlier images, and it drew something else obligingly. But I probably should have backed up and removed that from the context, because a few images later:
I don’t think that’s a realistic amount of space around the driver’s seat of an Audi A3 Sportsback (the model of car it invented for her to drive) that would let her sit that way. Probably an artifact of of its inclination to always want to show the character facing the camera.
So I saw this post in another thread:
And I told ChatGPT to make that shirt. And as it slowly revealed, I was thinking “yeah, yeah, good job following the sample pic I gave you…” and then… facepalm.
Reminds me of a Facebook group I see often called Terrible Maps (or something like that).
Saw this on Facebook just now, obviously made with ChatGTP
(I see lots of ChatGTP comics, but most of them are terrible both in humor and in execution. This one I think is pretty great.)
Ok, someone has to finally say it: for the love of god, it’s ChatGPT!
I was obviously referring to the Guanosine triphosphate version.
(In my defense, I get it right more often than I get it wrong.)
This one made me think of a typing test.
That might not be AI-generated. It might be a page from an actual comic, where someone just edited the speech balloons. Those have been around for ages, long before AI image generators.
I’d have a real hard time getting that particular panel layout in ChatGPT. It really loves doing regular tiling rather than other layouts of multiple images. I also suspect ChatGPT’s censor will kick in hard at any attempt to draw Superman or Batman. I think Sora might not let you draw them either; I’ve seen complaints in their feed about not drawing superheroes.
It might not be, but it was posted to an AI-specific group, and has ChatGPT’s characteristic yellow tint.
Now granted, I’m not a comics guy, but I’m having trouble picturing a comic where Batman and Superman are laying out in the grass staring up at the sky like a couple of teenagers…
this is the closest it let me get to the characters
I did some digging around earlier and it doesn’t look like Kerry Callen has made a new “Super Antics” in a long time, and the style doesn’t match anyway. And an image search doesn’t find panels from that anywhere else. I haven’t been able to refind the Facebook post, but it does seem to me that’s a ChatGTP creation somehow.
Midjourney v.7.0:
I’m counting at least 6 fingers on Superman in that first one. The first looks like Christopher Reeve, the second is somewhere between Dean Cain and Henry Cavill, and third is somewhere between Brandon Routh and the new guy.