Wednesday 28 February 2024

FEBRUARY 2024 -- SUMMARY



State of AI ART

AI ART shifted into some odd territory in February.  Most blatantly Google's AI refused to generate images of white people with it's new GEMINI AI program.


MIDJOURNEY

My issues were more personal as MIDJOURNEY refused my payment again (I think this is a "Stripe" issue). I like Midjourney because it allows me to "BLEND" two images to generate something entirely different. At least I got to try out Midjourney's new Style Reference feature, which they released about February 1st.

The irony is that I get less and less satisfaction as the AI Art programs get more and more powerful.  It seems that they are all pushing for the same archetype -- an anime version of Taylor  Swift -- and away from my drawing style. I noticed this agenda long before Google refused to generate images of white people.


STABLE CASCADE

I found another program that will "blend" images -- Hugging Face: STABLE CASCADE.  Moreover this program will accept my figure drawings (whereas Midjourney often rejects them for being NSFW).  However the results are not very good when I "blend" two of my raw figure drawings together:



Stable Cascade "blends" better with more robust images (but still not as well at with Midjourney), but does much better with text-to-image prompts:





LEONARDO.AI

I like LEONARDO.AI best for translating images into my drawing style, when I use the model "KrrrlLong" based on my long drawings.

The best results come after I upload a photo or robust image into Leonardo.AI, and it transforms that image into my drawing style:



However Leonardo.AI doesn't seem to work as well when I upload one of my drawings, and try to alter that with "KrrrlLong" (the Leonardo.AI model based on my long drawings).  I can generate something interesting, but it used to be a lot easier in the past, before they started charging for the more powerful features:



With "Alchemy" (a more robust Leonardo.AI model) I can transform my drawing into a Bernini sculpture.  Then when I "refry" or upload the robust Bernini results, I generate better figures from the model based on my drawing style:

Bottom: The top image uploaded and "refried"
from a model based on my drawing style

Also the AI Art YouTubers -- like Matt Wolfe, Curious Refuge -- and  are coming out with more and more content, but 










February 28, 2024

 x