AI Image Generators

I like to see examples of using a Generative Adversarial Network (GAN) to create images. This Person Does Not Exist, for example, features realistic AI-generated human faces. And Janelle Shane’s “AI Weirdness” Web log is consistently entertaining.

The more powerful systems are not available to the public, though as of this writing I’m on a waiting list for one that I’d love to try (DALL·E 2).

But there are more limited applications that can be used by anyone, and they get better all the time. Here are a few of those. For each, I’ve included examples of things that I’ve entered when I tried them.


Pix2Pix

This application enhances line drawings. However, I think it’s no longer available because it became too popular for the host to afford.



This one produces abstract images based on text prompts. For example:

“a monster coming through your bedroom wall at night”


“a man robbing a candy store”


“who actually greets you when you get to heaven”

This one produces slightly less abstract images. For example:

“a shopping mall flooded with oatmeal”


“fish swimming in cheese dip”


“the most beautiful skeletons travel”


This is a limited version of DALL·E. It also uses text prompts and I find it impressive. For example:

“Robocop fights Ed-209 in a Hieronymus Bosch painting”

“Katamari Damacy as a Katsushika Hokusai woodblock print”

“Katamari Damacy as a Junji Ito drawing”

“Katamari Damacy as a Roger Dean painting”

“Dark Souls birthday party”

“Silent Hill water slide”

“The Exorcist Saturday morning cartoon”

Tags:  

2 Comments

  1. Great to see you playing with these tools! This whole area is a fascinating (and natural) extension of software development over the last few decades.

    ()

  2. I’ve been trying to push the limits of Dall-E Mini’s capacity for surrealism. More of my experiments:
    http://autofish.net/clysm/art/ai/dall-e_mini/index.html

    ()

Add a comment