I haven't posted in a while cause I liked doing big posts on Tumblr.
Been working on a new series of proxies using updated versions of an older frame of mine and a public-domain/stock-owned trained generator. The idea is focusing on what I am actually good at which is graphic design and art editing. Personally, I like the results. We have an Art Nouveau-inspired set, we got a Kaiju 80's anime set...
Hell we even got a set of lands inspired by the gameboy! So, this is *not* real pixel art. This is the part of the card I find interesting, the most. It is the result of a pretty complex script I wrote for Photoshop that takes a (rather poorly done) AI pixel art (in full 32 colors) that is about 8 times the size of this, uses an algorithm to try to correct for common pixel art errors, then resizes it (dithering patterned) to the right size for the GB (but a bit wider), rescales it again (NN) to 4x size, applies color filtering to the GB 4 colors, then adds in a faux pixel gird (its not perfect). I edit it before the grid for mistakes/errors. That image (below) is filtered onto the screen of the token PSD. The land as a whole was made in Illustrator, and Photoshop, and used Firefly (I have a commercial license) for the stock image pixel art.
Also doing some tokens :) Been a busy last two months of work. This shouldn't be a surprise but most AI "art" is really poorly composed and riddled with errors. A lot of them I can edit, paint over, etc, but there's a lot I simply cannot. But it is also obvious which stuff I did earlier and which I did recently; as Ive gone at it It has gotten so much easier to do larger re-paints and edits. So I'm hopeful.