AI text to image generated art 🎨

[AArdvark]

I just counted the arms on the "octopus"

Huh.

[RealNC]

santa.jpg

[RealNC]

vader_bunny.jpg

[Ice Cream Jonsey]

Darth looking about 5'1" here.

[AArdvark]

The flying pink rabbit COULD be a Sith lord, couldn't it?

"What is thy bidding, my bunny?"


THE
DARTH LUPUS
AARDVARK

[RealNC]

ice_cream.jpg

[AArdvark]

RobB, have you ever thought about becoming a rapper?

[bryanb]

Happy Valentine's Day, everyone!

[Casual Observer]

It would be awesome if we could include the prompts used to get these. I've been making AI images for LinkedIn posts, its fun.

[RealNC]

It would be awesome if we could include the prompts used to get these. I've been making AI images for LinkedIn posts, its fun.

Not even the prompts are mine. I use ChatGPT and ask it to make a DALL-E prompt for me. So basically I have an AI generate a prompt for another AI.

The future will be interesting...

[RealNC]

pizza_crime.jpg

[Tdarcos]

Sometimes it gets part of it right, and the models and resources they train it on can produce different things. Consider this:

a man and a woman, sitting on a plaid blanket, eating sandwiches above a manicured meadow, bottles of soda beside each of them, a wicker picnic basket of food in front of them

Craiyon (https://www.craiyon.com/) produced these:

Let's see, All of them got the meadow right, all of them have a picnic, and that's it. I'll number the pictures from top left to bottom right 1 - 9, and describe what's wrong.
All missed the sodas.
1. No basket, otherwise this is a pretty good picture.
3. Basket behind them, blanket is green check, not plaid.
4. Only one person.
8. Blanket has a green "check" area in the middle instead of all plaid.
9. One person, blanket is green check instead of plaid.
2,4,5,6, and 7 are very close and would be a reasonable picnic picture.
Also, the picture file sizes are tiny, none is more than 20K, and most are 14-17K. On the other hand, the screen capture I made takes 1.5 meg.
Let's see what #2 itself looks like:

[Da King]

4. Only one person.

In all fairness, that one person DOES look like a combination of a man AND a woman.

[Tdarcos]

4. Only one person.

In all fairness, that one person DOES look like a combination of a man AND a woman.

You're right, #4 looks like a woman with a man's face. So, we have a song for that:



However, #8 doesn't have even that weak excuse. It is clearly a woman, and only one person. I missed that error.

[RealNC]

Well, this is certainly something that arrived way sooner than I expected: https://openai.com/sora

[Tdarcos]

Philip DeFranco said it exactly on one of his broadcasts: Remember, this, right now, is the worst quality results AI will ever generate, and it will only get more sophisticated in the future.

I am also reminded of the warning given in the game Doom about the Cyberdemon: Be afraid. Be very afraid.

[Tdarcos]

I just thought of something else. Someone is going to use this to create customized pornography. Create your own custom ordinary, unusual, depraved, or kinky sex video you want, on demand. Let's also remember, the U. S. Supreme Court has already ruled¹ that pornography using simulated child images, i.e. virtual child pornography, is constitutionally protected by the First Amendment.

- - -
¹Ashcroft v. Free Speech Coalition, 535 US 234 (2002).

[Tdarcos]

Here is the same prompt❶ given to Open AI (https://openart.ai/) and this is what it generated:



I would consider either of these spot-on. Amazing. If I had seen these before I heard about this I would have thought they were photographs. Especially, examine the faces: flawlessly real. No uncanny valley here.

- - - -
❶ A man and a woman, sitting on a plaid blanket, eating sandwiches above a manicured meadow, bottles of soda beside each of them, a wicker picnic basket of food in front of them.

[pinback]

I'd plow her silly!

[Tdarcos]

I'd plow her silly!

That is a meadow, not a... Oh wait, that's not what you were referring to.

Never mind!