Donno. You have to be a paying customer to generate several images at a time.
Which I am, and I tried. I varied the list, adding my own background ideas, and it always picked the first one and ignored the rest.
WAG: Someone cut and pasted it from another client that did allow batch prompting?
I use it to make alternate video game covers to use in Steam.
“1940s aesthetic, private investigator, noir style game”
Then I asked it for another take in a voxel graphics style
I really just gave it the title and it more or less spit out this. I tweaked it some but this is the one I was happy with.
“Pool hall in Tron 1982”
“Generic Hitman game cover”
This was “dick york and dick sargent standing in the stephens’ living room doing the “two spider-mans pointing at each other” thing” It definitely got closer on York than it did Sargent.
Another small issue I’ve noticed with ChatGPT 4o is that it seems to not quite understand how roads/hallways work. I’ve tried to make two or more characters meet face to face and it will have them face to face across the width of the road/hall, when I think most people would assume that you would meet along the length. I still get the problem even when I ask for a profile view.
Here are examples of what I mean, generated from text-only prompts. The scenes/characters are overall pretty great, but I wanted the meetings along, not across.
I see that it likes repeating numbers on pool balls, too.
The bar code isn’t readable. Now I’m wondering if that is a thing, if it can create a readable bar or QR code for a specific request.
I just asked it if it could. It replied:
Yes, I can create readable barcodes and QR codes!
They’ll be fully scannable — not just pictures, but real working codes that you can scan with a phone or scanner.You can tell me:
- What kind you want (QR code, UPC, Code128, etc.)
- What it should contain (a URL, text, numbers, etc.)
- If you want any special options (size, error correction level for QR codes, etc.)
Want to try one? What would you like the code to say or link to?
And, to be fair, I’m not sure the barcode on the back of Agent 47’s head in the actual game and marketing material is readable, either.
Oh, how about illustrated thread titles? I did these a few days ago.
Moe Howard and the Australian Mermaid
Raiders of the Lost Parachute Pants
Summary
A pile of socks of various colors and designs lying scattered in a corn field. Most are gathered in a centeal pile, but a few loose ones lie scattered about the whole scene. Also include a few small plastic troll dolls lying in random orientations.Create a realistic widescreen Kodachrome dslr photo with narrow dof.
I think in this case I like the Gemini version better. More creative on the sock designs.
I just noticed: “Straight Dope Message Beard”.
I love how you brought life to my weird off the cuff thread title.
Here’s the prompt that I used:
I want to see a scene from Indiana Jones as if it was set in the 1980s with 1980s fashions. Indy should be wearing a red leather jacket like Michael Jackson used to wear and parachute pants like MC Hammer used to wear. He is wearing a white sweatband/headband instead of a hat. He is still inside an ancient stone structure examining a small gold statue on a stone plinth. Widescreen Kodachrome DSLR image with shallow dof.
I just copy/pasted that in to 4o. It gave me this:
Your Harrison Ford is definitely better but my statue is a little closer I think. I can’t quite place who my guy looks like.
The one I did was through Sora. (Which of course uses ChatGPT 4.o itself.)
The single most annoying thing about ChatGPT/Sora to me is the seeming randomness of who and what is censored. Not only are different sets of things censored in Sora vs ChatGPT but it can change from one prompt to the next in the same session of the same interface. And even different rules for different characters in the same franchise. I can get Han Solo, for example, but neither system seems to allow Luke Skywalker. You can get Disney princesses, as can be seen in some of those x watching TV images, but neither would let me make Snow White and the 70 Worfs. It goes on and on…
I tried for that concept in DE3 months ago when the new movie was still being talked about. Most of the renders got Worf pretty wrong, but looking back in the set of images, some were better than I remembered.
