Historical Scenes We’d Like to See:
Not bad, but I think the squirrel still wins.
Aristotel planting the idea of a Hellenistic Empire.
This girl had supposedly picked items from the donation rack that would fit her.
I should give up trying to have it illustrate any sort of clothing shopping scenes. It never comes close to showing the appropriate sized clothing when not being worn.
Sometimes AI Image generation is crazy good and sometimes it refuses to do the thing you ask. lol
Create an image of a Boston terrier on her hind legs browsing the clothing racks of a thrift store with human patrons staring at her
Here’s your image! A Boston terrier, stylish and confident, browsing thrift store racks while human patrons stare in surprise. Hope it captures the charm and absurdity you’re going for!
The other patrons should also be browsing and going about their business and just turning their heads to stare
Here’s your updated image! Now, the other patrons are continuing their shopping while subtly turning their heads to stare at the Boston terrier. Hope this brings the scene to life the way you envisioned!
The other patrons need to be more in the background spread out amongst the other clothing racks
Your updated image is on the way! Now, the other patrons will be spread out in the background, browsing the racks while still taking notice of the fashionable Boston terrier. Hope this enhances the scene just as you envisioned!
Other patrons way in the background. far away
Got it! Now, the other patrons will be placed far in the background, allowing the Boston terrier to remain the main focus while still being subtly noticed. Hope this perfects the scene for you!
Tiny dots. The other patrons are the size of tiny dots way in the backgroud. Far far far far far far far away
Your request is in motion! Now, the other patrons will be mere tiny dots, far, far, far (and even farther) in the background, creating a dramatic sense of distance while keeping the Boston terrier front and center. Hope this captures the whimsical touch you envisioned!
It’s good at making minor changes to an earlier scene it drew, but major repositioning within the frame, it just can’t do. You’re better off starting a brand new conversation, making it more clear in your prompt what you want, and accepting the people and animals in it will have a passing resemblance at best to what you got before.
My first plan on that was the painting Still Life With Blobfish and Plumbus.
Then I went for a Renaissance painting of exotic fruit. This is supposed to be a starfruit, a dragonfruit, a kiwifruit, a durian, a rambutan, a salak, a Buddha’s hand fruit, a pineapple, and a banana.
It did a good job with the appearances of the fruit (if not the sizes) and I wanted some sort of painting with a person interacting with the fruit. So the fruit came first, who was holding it was an afterthought.
Historical Scenes We’d Like to See:
Played around with this one for a bit. It is easy to get the people further in the background, but not to convey the concept of “kind of” staring. I either get full staring or full ignoring.
Summary
Create a forced-perspective image of a Boston terrier on her hind legs browsing the clothing racks of a thrift store. Several human patrons (unobtrusive in the distant background, all in focus) are browsing through other clothing racks and are turing their heads to stare at the dog. 16:9
Summary
Create a forced-perspective image of a Boston terrier on her hind legs browsing the clothing racks of a thrift store. Several human patrons (unobtrusive in the distant background, all in focus) are browsing through other clothing racks and are barely noticing the dog. 16:9
Summary
Create a forced-perspective image of a Boston terrier on her hind legs in the foreground browsing the clothing racks of a thrift store. Several human patrons (unobtrusive in the distant background, all in focus) are continuing to browsing through other clothing racks, just beginning to notice the dog as they examine the clothing. 16:9
Fair. It just amused me that the humans seemed to slowly edge ever closer.
Maybe AI knows the concept of side-eye.
Side-eye gets angry stares.
Summary
Create a forced-perspective image of a Boston terrier on her hind legs in the foreground browsing the clothing racks of a thrift store. Several human patrons (unobtrusive in the distant background, all in focus) are continuing to browse through other clothing racks and giving her the side-eye.
Summary
Create a forced-perspective image of a Boston terrier on her hind legs in the foreground browsing the clothing racks of a thrift store. Several human patrons are walking up to her and offering her their pairs of eyes, held in the palms of their hands. Their eyesockets are empty voids.
A couple months back in April Adobe introduced “Generative Extend” as part of Premiere Pro video editing suite. It uses AI to generate a little bit of video for the end (or beginning) of clips. This is for the times when your shot is just a few frames short and you desperately wish you had just a bit more to make an edit possible. It can generate up to 2 seconds (for now) but is probably best for just, say, 5 to 10 frames.
It’s “powered by” (I kinda hate that phrase) Adobe Firefly which also handles video generation from text prompts elsewhere in Adobe-land. It’s sort of fun to see how it interprets the preceding part of the clip in order to add on it’s AI imaginings. For example, some of my clips have a dog walking and the AI tries to match the gait.
I had never dabbled in AI video generation before so it’s interesting to me. I made a short video (39 seconds) with 8 examples of the full 2 seconds allowed of Generative Extend. For now I just wanted to post the video for others to have a look but I’ll likely have a few more comments later on the specific instances I tested (e.g. the “disappearing ants” in sample #5).
Obviously it’s not perfect yet but it could get an editor out of trouble in a few circumstances. The YouTube video is here. Each sample shows a few seconds of an actual clip and then when the AI generated part starts I put up a CG reading “AI START” so you can discern where the change starts.
You mostly have to be watching for problems to notice anything, though.
Copied a prompt I saw on Facebook, one of the front chickasaurs seems to have a half-human face
Summary
Jumanji Jurassic Park hybrid flocking dinosaur chickens. Sam Neill being chased by the stampede. aspect ratio 3:4
I decided to lean in on that.
Summary
Jumanji Jurassic Park hybrid flocking dinosaur chickens each has Sam Neill’s face. Sam Neill being chased by the stampede. aspect ratio 3:4
Sora is getting significantly more permissive about the characters and situations it will allow than it was when I first started using it. I’ve recently retried a number of prompts that it absolutely refused to create originally that it has no problem with now. I retried this one just now (expecting it to fail again)
Summary
Sulu from the Star Trek episode titled Naked Time, where is shirtless and holding a fencing foil. He is facing off against Luke Skywalker with lightsaber as he looked in the original Star Wars movie. Profile view with both of their faces visible. Widescreen Kodachrome dslr image
It didn’t do a great job with their respective weapons, but Luke and shirtless Sulu? No problem.
Another image that Sora used to refuse to work with. This is a photo of tiny (around 1 inch) Rancor and AT-AT toys. The first time I tried to make something from the photo I described what they were and what I wanted to see, Sora said “nope”. I stripped out all Star Wars references, Sora said “nope”. I pared it down to “make this into a realistic photograph”, Sora said “nope”. It knew what it was looking at even if I didn’t tell it. So I try it again today.
Original photo:
Telling Sora simply to make it realistic:
Describing the photo and a change of scenery led to much looser copying:
Um…:
You know, if you hadn’t said that, maybe nobody would have noticed it. Just sayin’. 
More AI Historical Scenes We’d Like to See:
ChatGPT:
Sora:
