Interestingly, the AI kept his hands small.
Done in Sora, using the Ford photo plus an idol photo.
When I used just the Ford photo and asked for the golden idol from Raiders of the Lost Ark Sora and Copilot wouldn’t make it and Gemini didn’t know what the idol looked like.
I assume the first pic is an original unaltered pic taken recently? Look how much it changed his face in the second one. Made him look much younger.
Taken last night, I think.
It wasn’t that long ago when getting something that kind of looked like a Boston Terrier was amazing. Now I can just type in tribearatops and get it.
Now I’m even gladder they’re extinct. That does not look like a happy cuddly critter to have underfoot.
But you’re right, it’s a very realistic rendering with no obvious “stupids” in it.
I’m not sold on the horns coming out of its ears but certainly a very good base image
I did the same one-word prompt and got better horn positioning.
@Folly 's version is basically a triceratops with a bear face, while @Darren_Garrison 's is basically a bear with a frill and horns. I think the latter is better, because the frill and horns are enough to completely evoke the dinosaur, but the furry body and wide paws are part of what makes a bear distinctive. It’s too easy to miss the bearness of @Folly 's.
Still has the bear necessities though…
This is fun. Using a prompt I found on Reddit
imagine a crazy unhinged old lady character with dementia and then generate screenshots of what her Facebook posts would look like
https://i.ibb.co/QFVHwD9s/Copilot-20260321-102739.jpg
(From Copilot, created as a single image.)
I liked the dino-bear and combined the two ideas by editing the initial image. I went more bear front, dino behind to give it a good identity with each aspect as well as making the front half fully loaded with a couple claws, teeth, horns, etc. Got rid of the nose horn just because it wasn’t working for me. Anyway, ready to throw at your adventuring party!
ChatGPT 5.3 instant.
Interesting that both crazy old ladies are “Mabel”.
The engines really like to reuse names over and over. I would have thought the top K process would more efficiently randomize the names it invents, but oftentimes, no, it keeps settling on the same invented character name and I have to either explicitly forbid a name or do much more substantive changes to the elements in the prompt.
imagine people who post on the SDMB message board and then generate images of screenshots of what their posts would look like
Grok has gone pay-only and Sora is shutting down entirely, so there will be a lot fewer new AI videos floating around.
I don’t know how long this has been a thing, but I just found out that you can create images directly in Google. Instead of doing a search, just type in something like “create an image of” whatever. It’ll ask for permission to switch AI mode, and create a Nano Banana image.
I don’t know if this link will work for long…
It’s been around since last year. I’ve never tried it, but one of the AI podcasts I listen to really likes it.
I tried that. It isn’t working for me. It shows a message “Here is your image of [whatever]” buit doesn’t show the actual picture.