wektor

Well-known member
papal fallus.png
papal fallus
definitely a power electronics project
(note this is not really a fallus and it is growing out of the arm, so hopefully this will not be deemed NSFW, it's just a bit kinky)
 

william_kent

Well-known member
recent developments since the release of Stable Diffusion

people have discovered the Stable Diffusion img2img script: "make variations of input image"

here's a blog post showing how someone used img2img to transform their untrained ham-fisted ms paint image into sci-fi dystopia

there is also diffuse the rest

draw a crappy ms paint style picture, supply a prompt and it'll generate improved versions

the image below shows my badly drawn penguin on the left and on the right we have the AI version...

1661940737455.png

prompt: distended penguin musician


and

Stable Diffusion Animation

given a start prompt and an end prompt the program will try and interpolate between the two - it's early days, the resulting animations seem a bit clumsy

plus GIMP, Photoshop, and Blender plugins under development
 

william_kent

Well-known member
Midjourney seems to produce the best Cthulhu clowns

1662034882601.png

cthulhu clown, photograph by hans bellmer, in the style of H R Giger, hellish, artstation, calotype, multiple exposure, photoillustration, biomorphic, ambrotype, hand tinted

1662035099578.png

insane cthulhu clown posse, photograph by hans bellmer, in the style of H R Giger, hellish, artstation, assemblage, calotype, multiple exposure, photoillustration, biomorphic, ambrotype

1662035229648.png

nendoroid of sad cthulhu clown, figurine in the style of H R Giger, figurine in the style of hans bellmer, hellish, horror
 

william_kent

Well-known member
AI wins state fair art contest, annoys humans

A synthetic media artist named Jason Allen entered AI-generated artwork into the Colorado State Fair fine arts competition and announced last week that he won first place in the Digital Arts/Digitally Manipulated Photography category

he used Midjourney to generate the image which won the category - judges were reportedly unaware that the piece was generated by AI
 

william_kent

Well-known member
another prompt engineering resource: an implemation of CLIP retrieval

Clip retrieval works by converting the text query to a CLIP embedding , then using that embedding to query a knn index of clip image embeddings

or in another words, search for what you're looking to generate, and see what image-text pairings the AI was trained on ( this quizzes the LAION dataset which was used when training stable diffusion )
 

droid

Well-known member
The current test SD implementations on MidJourney are the best tool available right now. Kicks the shit out of normal SD and Dall-E.
 
Top