Tuesday, 13 September 2022

AI Art Catch-up

My head is spinning from trying to keep up with all this new AI Art firepower:

AI Art that evolved from:
"Two aging hippies at McDonalds
in the style of R Crumb"


This is all happening so fast. In the last month I received access to both DALL-E 2 and Midjourney, though others had access to those programs a bit earlier in the summer.  I also signed up for Dream Studio Lite, which is fueled by Stable Diffusion.



The AI Art Programs:
All of the above cost money, after a trial period.

°

Hugging Face:

3D AI



AI Art Search Engines

Lexica is jaw-dropping.  It's a search engine of a lot of the Stable Diffusion AI Art creations, complete with prompts.  You can even upload a URL of an image and it will search on that (with better and worse results).  I wonder if all these images are free to use, suspecting that there is no copyright on AI Art.

Lexica


DATABASE  Search Engines
 
HOWEVER: One of my images did come up in the search results of "Have I Been Trained" when I entered the URL of one of my drawings on my blog:

My image in the results it the
third from left, tan paper...
This is what I got uploading the URL
 of another one of my drawings



May 8, 2019
October 13, 2012
I entered the URL of the image on the left,
and one of my drawings (on the right) 
surfaced in the results

I uploaded this drawing to the 37th World Wide SketchCrawl, which apparently makes up part of the database the GAN was trained on.  It looks like it "trained on" at least one of my drawings 😀!



SHOWCASES

Some people have already tried to make something more substantial from AI Art.  For instance, a lot of people are making YouTube videos illustrating the lyrics of particular songs:


King Crimson
illustrated by Midjourney


Other people have made AI Art online comic books:


My AI Art Creations





AI Art made in Dream Studio Lite and DALL-E 2



Too Much Covfefe



My Tedious Life




My Previous AI Art Projects



PRINT OUTS

Weston printed out some of my AI creations from last year, and they came out very nicely.


High quality ink jet prints


Weston's archival ink jet printer did a lot better job than using the oversized printer at FedEx Office:



I wanted to framed some of the prints in cheap 8x8 inch frames from Walmart and have a pop-up show:



We already have had a few framed 8x8 inch pop-up print shows, and they look  nice:

Framed 8x8 inch prints
at the Los Griegos public library


However after seeing the sheer volume of decent AI Art on Lexica, I had to pause my idea of having fake AI Art pop-up shows.


CONCLUSION

Many of the AI Art creations are phenomenal.  However it is still hard to get exactly what you typed into the prompt.  DALL-E 2 tends to deliver literal and dry results.  Midjourney gives more artistic results, but tends to generate images that would promote a Tim Burton movie.  Stable Diffusion seems to allow more artistic nudity.

Of course there is a lot of bickering amongst the artists, especially after Jason Allen won first prize at the Colorado State Fair for an image that was generated by Midjourney.  However the "now" with AI Art has to be similar to when the artists first confronted photography back in the 1800s...artists just  have to figure out how to "pivot," as my friend Robert Atkison characterized this moment.


"Unusual drawing style" --
Maybe not that "unusual,"
the programs are always defaulting to images of women


Midjourney
DALL-E 2
DALL-E 2
"Running skeleton in  claymation"


CENSORSHIP: DALL-E 2 reprimanded me for typing in "horny toad" or "gringo," though it didn't object when I imported an image with a "horny toad" in it, made in a different AI program:

ArtBreeder Collage
DALL-E 2
DALL-E 2
"a horny toad riding a javelina"



Other links:


AI HOLOGRAM
backwards experiment

I made an hologram the backwards way.  

1)  First I asked Midjourney to generate a "black and white depth map with gradual gradations."

I choose to work with the first image


2)  Then after I chose and enlarged the first result, I imported it into Deep Dream Generator and applied a colored style to it.  Specifically I applied the style of a section of the paint pour image below:


I cropped the center of this image
and used it in Deep Dream Generator
as a style to transfer to other images


3)  I created an image of the result next to the original "depth map" using Photoshop Elements, and imported that into the Looking Glass Portrait display.  The 3D holographic result was pretty shocking -- download it from THIS LINK:

Import this image
into the Looking Glass portrait display
for a strange holograph




Posted by at

No comments:

Post a Comment

Note: only a member of this blog may post a comment.

Subscribe to: Post Comments (Atom)