AI-generated images thread

[Argent Stonecutter]

So I needed a dingo tucking into a plate of veggies and got this.

So I thought "that's a very Australian backdrop, what if I made it a coyote instead?"

It's dark, but that does look more like the US southwest.

I tried a jackal but it kept showing me interior shots with a blank white background. It took adding "campsite scene" to get this.

 

[Dakota Tebaldi]

lol, AI really does like fires in tents doesn't it

 

[Argent Stonecutter]

Various programs doing "a fox in a beret and a poet's shirt taking a photograph of a butterfly on a daisy using an old polaroid camera".

 

[Jopsy Pendragon]

GrrrrRRRRRRRRRrrrr....

Really struggling to get this to turn out the way I want it to.

If the goatee is right... it ages the face several decades...
If the face is right, the ears are wrong...
if the ears are right, the hair becomes a mullet...
if the hair is right, the face turns green...
If the face is right, the goatee vanishes...

/nghhhrhrhhrhghghhgh!!!!!

Well, the hair color is all washed out, but I suppose this is 'close enough' to my typical/signature Second Life avatar's appearance:

 

[Bartholomew Gallacher]

A muddy Trump.

Spoiler

 

[Khamon]

if the ears are right, the hair becomes a mullet...

ha ha ha, ha ha ha ha ha, sorry but ha ha ha

if the hair is right, the face turns green...

Hey! Hey now! Hang on!

Does it help if you broaden the nose and chin a bit. You can do that and maintain the sexah Jopsy appearance AFAIC. The features might not overlap as suddenly on a slightly wider face maybe.

 

[Jopsy Pendragon]

Hey! Hey now! Hang on!

HA! Yes, well -being- green works for you, I prefer to limit my verdanity to the keratinous bits.

I don't think it's the distance between facial features that's the problem... I think it's the training/sample data. For skin tone will dramatically impact facial features.. and will fight with hair color. Blond results in warm pale skin tones, Ginger/Burgundy hair often produces very cool almost clammy pale skin tones.

I've tried messing with chin descriptions, but it almost always does a belly flop into the waters at the bottom of the uncanny valley. (/shudder sooo creepy huhhghhgh!)

And when I try to twiddle the nose a little.. it's a whole bunch of "Scroll of Nothing Happens". Like the engine doesn't know any of the descriptive words I'm trying. "Button Nose / Snub Nose / Upturned Nose / Puckish Nose / fae folk nose / a nose like Dylan O'Brian or Meg Ryan / ..." ugh. Strangely enough, I -sorta- almost get the effect I want by adding "youngish Estonian male face" to the prompt. =D

I'm not committed enough to actually try to create a training data set or LARA or whatever, but clearly that's what it's gonna take at some point. When insomnia hits, I certainly waste enough time generating faces for fun, but not quite awake/alert/sharp enough to do more than throw random words at a image generator to see what happens though.

A muddy Trump.

Swampy Trump, aka "Swumpy the Tramp"!! =D

 

[Khamon]

You can write a whole chapter describing how you arrived at "youngish Estonian male face."

 

[Free]

Or maybe a book or 3.

 

[Free]

Ha!

We made a cat drink a beer with Runway’s AI video generator, and it sprouted hands

Gen-3 Alpha produces wild and whimsical results. Here's what it cooked up for us.

arstechnica.com

In June, Runway debuted a new text-to-video synthesis model called Gen-3 Alpha. It converts written descriptions called "prompts" into HD video clips without sound. We've since had a chance to use it and wanted to share our results. Our tests show that careful prompting isn't as important as matching concepts likely found in the training data, and that achieving amusing results likely requires many generations and selective cherry-picking.

Even sports the ubiquitous extra finger on its left hand.

can

It's a Trump cat: It needs both hands to hold the can!

 

[GoblinCampFollower]

 

[Argent Stonecutter]

I prompted a few engines with only the word "llamageddon", only Dall-E came though.

 

[Argent Stonecutter]

Gonna need a repair team.

 

[Argent Stonecutter]

New York Dingo:

Noo Yoick Dingo:

I don't know where that came from but damn it's cute.

 

[Free]

FLUX: This new AI image generator is eerily good at creating human hands

FLUX.1 is the open-weights heir apparent to Stable Diffusion, turning text into images.

arstechnica.com

So the AI porn won't keep creeping me out now.

 

[Khamon]

No, it will, still.

 

[Argent Stonecutter]

Night Cafe has Flux, I tried a few variations of the good old werewolf drinking a Pina Colada at Trader Vic's, his hair was perfect. This one came out as a were-scottie...

And this one is a dead ringer for Legoshi from Beastars:

 

[Argent Stonecutter]

This one, the prompt was Coyote Spirit creating Humanity and it's noble AF even if he's just posing for the camera:

 

[Argent Stonecutter]

 

[Argent Stonecutter]

Dingo Medic. Two quite different results.