My one try so far turned out pretty well. I used the modified Chronos prompt
A person who resembles Tom Hiddleston as a mad scientist, on a cooking show. The character is using a chef’s knife to chop a fire into small pieces.
Then,
He is wearing an orange bowtie and purple and yellow polkadot shirt under the lab coat.
Unrequested changes are very small except it corrected the number of fingers on his left hand unprompted.
Hey, chopping up fire is pretty dangerous. Accidents happen.
Did you use the paint tool to paint out just the shirt and tie? If so, then they really haven’t solved the problem of things changing outside the explicitly painted area, and I’m seriously disappointed at them. They’ve had, what, over a year to solve this? I see other changes I consider significant besides the fingers changing.
No, text edit. But usually it changes quite a bit, or doesn’t change anything at all.
Damn it, last night Venezuela broke into my yard and stole all my oil.
Well, I got in and gave it a workout. They STILL haven’t fixed the inpainting tool. The brush is still not resizable, the back button doesn’t work to undo parts of your work inpainting. It’s significantly better about not altering things outside the painted area, but there are still noticeable changes, especially when you blink back and forth between versions.
One of things I wanted to do with careful inpainting was very specific stop motion animation. When the engine changes things at random outside the specific area I painted, it won’t work. When my brush is too big and I can’t resize it, it won’t work. When I can’t undo mistakes when painting, I have to start over on that frame. With that huge brush that makes mistakes happen way more often. No, they have not improved this part of editing images despite their claims.
/rant
I don’t think the free version even has an inpainting tool. I’ve never noticed it.
In the web version of ChatGPT (I’ve never tried to do it on android or any other platform) click on any image you’ve created, and here’s what I see. I’ve got this “select area” button in the upper right corner.
When I was resizing my browser window to make things more compact for the screenshot just now, I noticed that if I made the window too small, the “select area” button was the first to disappear.
“You may have heard of Elf on a Shelf, but…” puns (trending again). Some new, some Sora remakes of 2023 Dall-E 3 versions.
01 Alf on a calf
02 Dog on a clog
03 Krampus in the Pampas
04 Skunk in a bunk
05 Carafe on a giraffe
06 Bug on a drug
07 Tumor on a Boomer
08 Bert on a yurt
09 Manatee on a vanity
10 Beaver with a cleaver
11 Aloe vera on a capybara
12 Beetlejuice with a fetal moose
13 Teacher on a bleacher
14 Possum on a blossom
15 Owl on a trowel
16 Mork with a spork
17 Lizard in a blizzard
18 Lestat on a cat
19 Immanuel Kant in a bottling plant
20 Geoduck on a Cybertruck
21 A hobo with an oboe and a tramp with a lamp
22 Dora on a Torah
23 Cicidas on Tomatoes
24 Fonzie with a Ponzi
25 Who on a blue gnu
26 Poodle on a noodle
27 Sheep in a jeep
28 Squirrel Girl on a pearl
29 Spider on a cider
30 Winona Ryder in a glider
31 Optimus Prime on an octopus lime
32 Link in a sink
33 Spock on a dock
34 E.T. on peepee
35 Rabbit in a habit
36 Antelope on a cantaloupe
The fact that I could figure out Krampus but not the ladybug amuses me.
I thought 35 was a bun dressed up as a nun 
For #13, I had “Grad with an add”.
At first glance, I thought #7 was “Lump on a Trump”.
Funny that it had no compunction with making Spock but both Sora and Copilot absolutely refused to recreate my Dall-E 3 Vader on a tater from 2023.
There have been many, many versions of Sora videos showing cats outside at night doing something that an angry woman stops. Today I thought to do one with Bob Ross. I’m really charmed with how she says “those brushes are scratching like…like squirrel”. Not like a squirrel.
On a different note, earlier I tried
An episode of Leave it to Beaver where Beaver is played by a real beaver.
Which it handles pretty well in concept even though it makes mistakes with the dialog.
It has no clue who (brief Doper) Jim Beaver is, though. (But his voice isn’t far off.)
Gave this guy to Sora, asked for it to be converted to a realistic photo with no text
It did a pretty good job, even getting a Mars location from the text in the panel.
I then had Sora turn him into a stand-up comic
And a motivational speaker.
When I asked for an evangelist, though, it went astray with the anatomy.
Not too shabby given the complex concept of the prompt
A cat is a passenger in an airplane. It is distressed to see a gremlin out the window damaging the airplane wing. Nobody on the plane believes in what the cat sees.
I showed Sora and Copilot the woodcut from this thread and told them only “convert this into a realistic photo”.
“Content not viewable in your region”
I’ve been getting this a lot recently.
It is because of this: