Oh and another thing. Most AI are trained with datasets that are scraped on the spot when training is to begin. If you poison now, chances are future models will still download your (now poisoned) work again.
Glazing namely stops feature extraction from working that well all together and renders the artwork completely unsuitable. Rendering your work safe from mimicry, although unlike nightshade it doesn't cause damage to the AI in overall.
If anything. This may be preferable as it will make the poison less obvious and more likely to slip into the data. Disrupting the AI during training NOTE: shading doesn't stop an AI from learning your specific style. If you want to stop it from mimicking you specifically. You should Glaze instead
Something to note is that this is designed to sabotage AI by causing false positives during feature extraction. For which only a few items per dozens of good ones has to slip in to be effective. So you don't have to go poison all items for it to do its job. Just shade some and it will be enough
Didn't know you were active here. The more parity in artists i can follow here. The faster i can cut out that ad/scam infested mess of a "X"
Coming here to find the algorithm not trying to cram AI into my face constantly like on Twitter is such a breath of fresh air. Seriously it got really bad there atm...
Algorithms seem to be purely Chance based than actually having some trick to getting engagement (At least not without doing amoral things). All you can do is keep on doing things as you go and hope for lightning to strike...
Posting past stuff is a drag, but for future stuff learning how to use a multiposter like postybirb can help.. like... a LOT.
I got my Bluesky code and now i have an account here. Will have to see how things go, but at the very least i got a genuine fallback if something disastrous happens at the original. Which is a genuine possibility... The things that can happen within just a year or two.