DALL•E Mini/craiyon - Description-based AI Image generator (that won't let you enter naughty things)

[Stephanie Bustcakes]

Woah, Momma!

 

[Metamorpho]

 

[AnOminous]

 

[Ether Being]

Got in that v2 service club (with 50 complementary tokens that expires within a month, go figure), but that invite waiting time was worse than getting into Midjourney's that Dall2 as the same issues others pointed out it's aggressive censoring(or risking get ban), credit token usage, and lack of trial and error requiring accurate prompt-fu(via sandboxing craiyon).
Kinda makes me wonder why there hasn't been an executable native version of these AI art generator programs with a small offline database, closest I've seen was Style2Paints using GPU accelerated TensorFlow processing just for auto-painting lineart.

 

[Least Concern]

 

[DeadwastePrime]

i got access to dalle's ai generator so now i can do stuff like this

 

[Buttz McSmellington]

Here are some abominations I made with Craiyon. I also got into the V2 club.

Spoiler: Craiyon

Spoiler: V2 Stuff

"TMNT Michaelangelo in the style of painter Michaelangelo"

"Junji Ito Corgi"

"Pablo Picasso Batman"

 

[Fougaro]

"I can't fool myself, I don't want nobody else to ever love me, you are my shinin' star, my guiding light, my love fantasy~"

 

[DeadwastePrime]

 

[Chao Garden]

why the fuck did it give me pictures of dogs and bugs lmao

 

[The Last Stand]

I recently got my DALL-E 2 invitation, and I am coming to the same conclusions.

Additionally, practicing the dialect is impractical because of the render credit system, or more specifically the lack of a system to reject misfires & refund wasted credits. If they are going to deliberately tamper with the system to return suboptimal results from carefully worded input, we shouldn't be punished with irreversible credit deduction as if it were a user error. Seeing the credits dwindling without any satisfactory results gives an unpleasant feeling compared with craiyon's trial-and-error freedom.

Same, received my invitation as well.

I learned the hard way that asking for variations for existing AI creations counts for one credit. I ended up wasting three credits on that feature.

I am impressed that it can handle decent results from a distance. It's not good with faces; they come out smeared.



a 1930 cadillac fleetwood driving up a gravel road on her way to a cottage

There's an option on Dall-E 2 to "edit" uploaded images but from my experience, it doesn't work. If it wasn't for the credit system; I'd be having more fun with it. For now, I have to be methodical with my inputs as to not waste any.

 

[Lake of fears]

 

[The Last Stand]

1930s car with a flat tire in downtown chicago



Jean Harlow wearing a zoot suit while next to a 1930s car in downtown chicago noir style

DALL-E 2 cannot do faces well. But the first one is fire.

 

[Captain Syrup]

 

[737 MAX Stan Account]

 

[whatever I feel like]

View attachment 3640616

All those Trumps have basically Trump faces, teach me your ways, AI-sama.

 

[Captain Syrup]

View attachment 3662041

 

[Mysterious Girlfriend X]

 

[Lake of fears]

 

[Captain Syrup]

View attachment 3662655
View attachment 3662657View attachment 3662658