I have recently started doing AI art. I have, in fact, been working with, on, and around AI for over a decade now. I've also spent a great deal of money on art school, over 40k in total of my own and other people's money. I learned how to use Illustrator and Photoshop, and how to use a Wacom tablet.
Most of my newbie abominations of anatomy and form and color happened when I played Second life, far longer than a decade ago, now.
Moreover, making an AI generate "something" is easy. Making an AI generate "something specific" is not easy. For me, currently, getting specific results is going to involve a lot of editing in Illustrator on outputs until they conform to specifications, and then sending the results back through to clean them up. It's a tedious process, just like any other art, and involves work, color theory, composition, and a good deal of anatomy in reconstructing things that get mangled or destroyed. I don't have a sub to Photoshop, and can't justify one unless I open for sale!
There is a difference. The difference is that my goal is to produce some manner of work product that does not just involve a number of pieces of art. My goal first goal here is to eventually be producing whole character embeddings and LoRA training sets that will faithfully allow a user to produce a character on demand in any pose, hopefully allowing transfer to other styles. The set is art whose value mostly exists as a set, but the model is the bigger piece of art in and of itself.
I also intend to help produce a workflow that will allow artists to make and produce art in their own unique styles. Personally, I really like this style and will continue using the tool in this way to produce pieces like I have been, and taking results I like. Perhaps eventually this selection process will create a style; perhaps that is all a style is, albeit one of "inside" rather than "outside". Only time will tell.
If I ever end up making real money doing this, that money will go to my hobbies and interests: commissioning traditional artists, buying tools for "material arts" such as gem cutting and jewel craft, paying sub fees for tools such as Photoshop and Illustrator, 3d printer resin, and the like. Maybe I'll branch out into sculptures some time? That would allow me to create embeddings for specific poses, too. There's so much cool work that can be done in this space, but it takes skills, and not just that but specific ones, to do well and to do great things.
I reserve the right to retain all images that my workflow ultimately produces, and use any LoRA training set images I produce for maintaining a master training set for running training epochs on "Furtastic" versions, to be dubbed "Cubtastic", as well as sets to create embeddings for the system. There will, hopefully, be a public version containing all the non-private images stripped of character tokens, but any character embedding or LoRA would likely continue to work with the core version.
Anyway, think what you want about AI art, but I see it as a force multiplier, where one person can make their game, their character art generator, their whatever, so we can have more of everyone's unique visions rather than seeing the world confined to just our own.
| Viewed: |
23 times |
| Added: |
1 year, 6 months ago
23 Jul 2023 05:50 CEST
|
|
|
