DALL-E Mini The AI Image Generator

awdrifter

Well-Known Member
Joined
Aug 29, 2004
Messages
3,447
DALL-E (stylized as DALL·E) is an artificial intelligence program developed by OpenAI that creates images from textual descriptions.

It uses a 12-billion parameter[1] version of the GPT-3 Transformer model to interpret natural language inputs (such as "a green leather purse shaped like a pentagon" or "an isometric view of a sad capybara") and generate corresponding images.[2] It can create images of realistic objects ("a stained glass window with an image of a blue strawberry") as well as objects that do not exist in reality ("a cube with the texture of a porcupine").[3][4][5] Its name is a portmanteau of WALL-E and Salvador Dalí.[2][1]

Many neural nets from the 2000s onward have been able to generate realistic images.[2] DALL-E, however, is able to generate them from natural language prompts, which it "understands [...] and rarely fails in any serious way".[2]

OpenAI has not released source code for either model, although a "controller" of DALL-E is available on OpenAI's website, where output from a limited selection of sample prompts can be viewed.[1] Open-source alternatives, trained on smaller amounts of data, like DALL-E Mini, have been released by others.[6]

According to MIT Technology Review, one of OpenAI's objectives was to "give language models a better grasp of the everyday concepts that humans use to make sense of things".[7]

Here's the link to the DALL-E Mini online version.

I asked it to generate "google lambda ai". Here's what I got. lol.
dallemini_2022-6-20_22-4-45.png


This is "silent hill nurses"
dallemini_2022-6-20_21-56-8.png
 
It's funny how I get a bit of a mental block when presented with something like this, I did come up with some things to try though.

BananaBoat.png
CheeseHorse.png
DalekPicasso.png
PencilDonald.png
WatercolourTeletubbies.png

Craiyon, which appears to basically be the same thing as Dall-e, came up with slightly better results for that one. Still the stuff of nightmares...

WatercolourTeletubbiesCraiyon.png
 
OK I'll give that a try.
WornOutFaces.png


Er... Well that's a bit creepy.

GhostRiders.png


OK that's cool.
 
Thought I'd push it on something it seems to really struggle with:

dallemini_2022-6-27_19-36-29.png
dallemini_2022-6-27_19-41-55.png
 
It...it's not wrong.


dallemini_2022-6-27_20-4-52.png
 
Last edited:
Might be too complex of a reason?
dallemini_2022-6-27_22-16-0.png
 
dallemini_2022-6-28_12-28-38.png
 
Found on imgur, thats actually good. :D

1656448800830.png
 
Some of these are really brilliant, those bread houses are great. Shame there isn't a way to help it learn with feedback on what's good and what's not. Speaking of houses, I also gave it a house challenge earlier.
ASCIIHouse.png
EtchSmiling.png

BlueprintHelicopter.png
 
Top