So, yeah, it picked up on “Bob Ross” from you and it didn’t want to contradict you.
Yeah, it definitely tends to try to agree with you, unless you’re blatantly, factually incorrect about something. Although I did honestly feel like what it produced strongly resembled Bob Ross’s work.
Ross was definitely my first thought from that image, too.
I’ve found that asking it detailed questions about the image it just produced only works up to a certain point. It’s 1000% better than Dall-E ever was, which is something, but it’s not all the way there yet.
I’ve got an original character that I’ve defined to have a specific appearance, to make them more recognizable in illustrations and perhaps more marketable. At least more memorable. I can get ChatGPT to draw them correctly from just the text description maybe 75% of the time. The odds of success go down the more other complex things need to be in the image, though. 
I sometimes quiz it about things it got wrong, and oftentimes it just fails to recognize that particular detail it drew doesn’t match the text description I asked for. Or it agrees it got things wrong, but it still can’t describe what it did instead.
I gave it a detail from a Bosch tryptic to “make realistic.” Nailed it.
Awesome.
Someone else’s prompt
Summary
A high-speed photo captures a green basilisk, also called the “Jesus Christ lizard,” sprinting across a calm water surface, desperately fleeing from a bullet flying just behind it. The lizard is mid-stride, with one leg lifted and water splashing behind it, showing panic and speed. The bullet hovers slightly behind and to the right, emphasizing the sense of danger. Concentric ripples trail from each footstep, and the otherwise mirror-like surface reflects both subjects. The lighting is soft, and the background neutral, enhancing the cinematic tension.
My revision
Summary
A high-speed photo captures Jesus Christ sprinting across a calm water surface, desperately fleeing from a bullet flying just behind him. Jesus is mid-stride, with one leg lifted and water splashing behind it, showing panic and speed. The bullet hovers slightly behind and to the right, emphasizing the sense of danger. Concentric ripples trail from each footstep, and the otherwise mirror-like surface reflects both subjects. The lighting is soft, and the background neutral, enhancing the cinematic tension.
Three out of four of the images in my version had Jesus running in the same direction as the bullet, racing it like The Flash.
Caption: “Damn time travelers! This isn’t how I’m destined to die!”
Interesting that they made the bullet so big, relative to Jesus.
Maybe that’s Mini-Jesus.
The time traveling meddlers wanted to be sure, but they didn’t count on Jesus’s other divine superpowers not revealed in the gospels. 
Inspired by a brief glimpse I had of a poster on Adam Savage’s workshop wall in a recent video.
I’m not very satisfied with this. I wanted a more realistic flying car, and I wanted it to look like the human bodies could actually fit in it, but after several attempts, this is the best I got.
What happens if you use Jetsons as a key word? Tho looking at this the Ai came close.
There actually is a company getting close
Yeah, I did use Jetsons as a keyword. The prompt was:
Make a photorealistic image of the four members of the Jetson family in their flying car. They should look like realistic humans but as close as possible to the cartoon counterparts. The background should show a futuristic cityscape but not so detailed that it distracts from the car in the foreground. The car should look close enough to the cartoon car that it is immediately recognizable, but update it somewhat to look more modern and realistic.
And then I made a few further tweaks to try to improve the image.
I’m not at all suprised that ChatGPT and Copilot refused to make Alice, Dorothy, Sarah, and Chihiro kicking a Peter Pan lying curled in the fetal position on the floor. What suprises me is that Sora didn’t.
Devin Stone, Pixar version (LegalEagle on Youtube)
Playing with the earlier Jesus prompt. I replaced the bullet with a robot character modified from a prompt I found on Facebook and changed the angle.
I discovered that ChatGTP doesn’t do a very good job of depicting a crucifix from a distance. This one looks like a mummified monkey corpse.
I replaced the fembot with something that makes more sense.
Wait, which one of those guys said “I’ll be back”?
A dingus with a plumbus. 35mm dslr photo with forced perspective and shallow dof. 9:16 portrait mode.
Ran twice in Copilot and one 4 variation set in Sora.
Changed “dingus” to “major contemporary celebrity”, got three versions with Kanye West, one with what Google Lens says is Justin Bieber. Added “female” to that, got three failed images and one Sydney Sweeney.
