Here’s a variation with Sora:
So that’s something more like what Baphomet would do? (WWBD?)
Bravo good Sir!
I’m not generally a fan of bumper stickers, but a “WWBD?” sticker would be totally worth it. Like a good custom license plate, it’d have everybody behind me wondering what the “B” stood for.
Apparently the full transformation process includes also changing the shape of the bottles.
In case you are wondering, I only had to replace “Jesus” with “Baphomet” in the prompt. No description necessary, it knew what to do.
That’s awfully thin-looking ragu.
never mind
This is the Jesus who came out of the apocryphal Infancy Gospel of Thomas.
I was going to run the Jesus prompt through KlingAI by itself, but decided to use my favorite image as a reference instead.
Paste that video URL in a browser to see the video full resolution. There are interesting details, like how the water moves in the bottles and how the cars on the Ferris Wheel rotate properly.
There it is again. There are computer screens in the scene. The context knows about the email that guy just received. Show content on those screens, no matter how little sense it makes!
This somehow reminds me of a fail pic somebody posted someplace taken 10, maybe 15, years ago. It’s an actual ad for a game console. It shows 4 professionally cute young Asian women holding game controllers looking excitedly towards our camera POV which is standing in for the screen they’re supposedly looking at. They’re evidently having great fun playing with this new whizbang console.
We also see the rear face of the game console device sitting on an adjacent table. With every exposed connector empty. No wires. None to the electrical power inlet, none to the video screen, none to the game controllers. This in an era where everything was wired. Oops.
The added fail caption was something like “Maybe they’d have even more fun if it was plugged in.”
Looking at this AI-generated workplace makes me think “Maybe they’d be even more productive if they could see what they’re working on.”
Saw an image from a catalog of children’s costumes in the 1920s. The “saucy” one stood out (apparently a brand of sauce in the UK). I couldn’t find a complete version of the toothpaste costume from the edge of the image, but did find a better one from (I assume) a later year. I brought them into the modern era.
Nitpick (because this is the dope), bringing them into the modern era would mean spelling “to-day” as “today.” :d&r: I have no doubt the AI would have done it if you’d asked.
A modern girl found that old ad and made one of those other costumes for Halloween:
I didn’t ask it to include the text but it did anyway. 
Here’s someone else with one of the costumes!
The above image was made on Sora using the original drawing as a reference image. This one was made through Copilot with three experiments. First, the prompt is text-only. Second, I removed the description that the skirt be plasticy. Third (and something I hadn’t tried before just now) I used only hex codes for describing the colors that I wanted. (The dress and dots are #B81D13, #EFB700, and #008450.) It got remarkably close to the original outfit, even the odd shape of the hat (which I described as “an unusual hat, black with a wide brim, tall and tapering to a flat top, like a truncated witches hat”).
I’m looking at what the AI shows you of the image in progress. Sora and ChatGPT don’t show in-progress images for me (though apparently ChatGPT does for Ponderoid) but Copilot does show them. And it is interesting to see that apparently the AI doesn’t work on the whole image at once but does it in segments from the top down, knowing exactly what the top will look like while the lower parts are just vague ideas. Like in this example. It gives three in-progress images. In the first stage it doesn’t appear to know yet that she will have wings. In the second it doesn’t appear to know about the arrangement of her legs, and in the third it doesn’t appear to know the arrangement of her feet. But earlier AIs that show progress do the whole image at once, refining the whole thing with each stage.
(That image is the result of asking for a realistic photo of this):
Something I noticed when trying 3 image splits is that the AI never made the divisions equal even when I specifically ask for it, and it is always the leftmost section that is narrowest and has details poorly framed and chopped off. Which seems to suggest that it allocates starting from the right and working left. But in actual rendering it does all three segments at once.
Social comment:
In the original 1920s catalog, the traffic light and artist’s color swatches & palette costumes are labeled as suitable for girls ages 11 to 16.
Somehow I have a hard time imagining a modern girl much over age 7 thinking that’d be a costume they’d be willing to wear.