Meet DALLE MINI, the AI bot that generates images based on your suggestions

This one, based on Franz Marc, I upscaled and cropped a little. Still has annoying bits at the edge that I'd get rid of, but the overall impression I found pretty astonishing

https://i.imgur.com/HdLf3ar.jpg

Played about with Matisse too. I think the prompt for the first was "A blue and white cut-out of a laughing face, made by Henri Matisse". The second one started the same and I forget the subject. Third was a self-portrait.

https://i.imgur.com/W2YtlOW.png

https://i.imgur.com/TksflcI.png

https://i.imgur.com/egoVYCc.jpg

Camilla and Charles painted by Gilbert and George, I think

https://i.imgur.com/ZqpqKoL.jpg

I like the mysterious figure haunting them from behind.

^ diana

These are great! I need check this one out.

A few from actual DALL*E

https://i.ibb.co/PFnvv9J/Screenshot-20220917-103709.jpg

https://i.ibb.co/GvPmFYg/Screenshot-20220917-102716.jpg

https://i.ibb.co/GMWHVxh/Screenshot-20220917-103605.jpg

https://i.ibb.co/mbyK9qJ/Screenshot-20220917-102342.jpg

https://i.imgur.com/PVKGLxr.png
Salvador Dali painting of Phyllis Diller

aaiigh!

is there a thread already for ai writing software?

here it is....

thank you

https://i.imgur.com/OeiZKql.png
a 3d rendering of Gustav Klimt's "the kiss" in the style of Bubsy 3D

https://i.ibb.co/f2cqWFC/Screenshot-20221005-191234.jpg

https://i.ibb.co/XZp7VbK/Screenshot-20221005-191003.jpg

https://i.ibb.co/R2K8nxx/Screenshot-20221005-185953.jpg

https://i.ibb.co/xSp5qnh/Screenshot-20221005-185853.jpg

https://i.ibb.co/RBLbW6d/Screenshot-20221005-185750.jpg

https://i.ibb.co/hL9H0pg/Screenshot-20221005-185553.jpg

https://i.ibb.co/yy13Hx0/Screenshot-20221021-161935.jpg

I am really amused by the AI's attempts to make words that are always nonsense words 'MOTSOOS"

Its interesting because IA can generate text and meaningful phrases but DALL-ED type AI has no idea what letters really are

the Ganfield Twitter feed is a ton of Garfield strips feed into a GAN and often produces interesting results but, yes, no understanding of text

10/21/2052 pic.twitter.com/xXP3vLzbC2

— GANField (@futuregarfield) October 21, 2022

https://labs.openai.com/s/HmbJOqCt4nyNz776uN26cSla

superman pope

in other news https://twitter.com/yeetgarf

haha

xp apparently it’s not nonsense but dall-e’s own internal language:

DALLE-2 has a secret language.
"Apoploe vesrreaitais" means birds.
"Contarra ccetnxniams luryca tanniounons" means bugs or pests.

The prompt: "Apoploe vesrreaitais eating Contarra ccetnxniams luryca tanniounons" gives images of birds eating bugs.

A thread (1/n)🧵 pic.twitter.com/VzWfsCFnZo

— Giannis Daras (@giannis_daras) May 31, 2022

have you tried feeding the words back in?

The "internal language" referenced in that tweet has nothing to do with the way DALL-E generates text within images. It has to do with how DALL-E interprets text inputs.

I took a look at the BPE encoding of the name DALL-E uses for birds.

Its "apo, plo, e</w>, ,ve, sr, re, ait, ais</w>".

Apo-didae & Plo-ceidae are families of birds, each with 100+ species. Apo-diformes is the biggest order of birds with 400+ species of birds. https://t.co/DpP7Rjq2Gf

— kluge (@BarneyFlames) May 31, 2022

this creature has adapted to the crushing pressure and oppressive darkness

https://i.imgur.com/shNQkTe.png

Ai is like dogs. We created a cute monster.

The "internal language" referenced in that tweet has nothing to do with the way DALL-E generates text within images. It has to do with how DALL-E interprets text inputs.

that's how i interpreted it, the internal language is what the prompt gets translated to before it's rendered, the interesting aspect is that dall-e leaks that into the the image itself.

i do get insects with "Contarra ccetnxniams luryca tanniounons", i tried doing variations of "two .. talking about .., with subtitles" and then using the generated text but just got some seemingly random images.

human pit bull brundlefly - kind of adorbs

https://i.imgur.com/NJwBakq.png

I wonder if dalle text is how Chinese companies come up with English company names. Like hntmao (actual name)

https://i.imgur.com/VnKtGn6.png

glad to see the Simply Red reunion tour is off to a smashing start

An Eldritch abomination descends upon a small Kansas town, Pixar style

https://i.imgur.com/ZSHLGE4.jpg

https://i.imgur.com/pRCzBGX.jpg

Tried to do something with Rum, Sodomy and the Lash but sodomy is banned

Rum, proctology and the Lash

this may be the weirdest result I've gotten

https://i.imgur.com/SYAG4k8.jpg

Midjourney could pump out the album covers for the fake jazz and ambient groups that Spotify creates to siphon listeners

https://i.imgur.com/lVNCjUy.png

And I could probably generate that music too

Ai not and I

You don't give yourself enough credit.

https://i.imgur.com/kgMr9FX.jpg

here's an AI colorizer which is astonishingly good, no signup or payment necessary

https://palette.fm

https://i.imgur.com/v2CE7bk.jpg

I tried this

This whole thread >>>>>>>>

Still from 1972 british comedy film "Help! I'm Trapped behind the Forth Wall", widely considered the funniest film of all time pic.twitter.com/mBGiXqsh6I

— Erika Chappell 🏳️‍⚧️ Professional Simpsonsologist (@open_sketchbook) November 5, 2022

holy shit. what????

further clarification if neede d the original "Bessie of the Opera" pix were black and white

https://i.ibb.co/xJhkGH9/Screenshot-20221113-230643.jpg

https://i.ibb.co/YBQ0kn1/Screenshot-20221113-232550.jpg

this is my favorite anti #aiart take now pic.twitter.com/VEzcbD69y7

— proxima centauri b ▫️ open edition live! (@proximasan) January 17, 2023