It's quite fun, and fascinating, trying to create images from text, though also somewhat frustrating. There's so much going on here, from the "unconscious" biases of the learning model (initially, I didn't indicate either gender or race for the image, for example) to its stubborness regarding some details - but still great fun to play with. I had a go at the same scene Bobbie experimented with, using getimg.ai, which does allow for some nudity. I then tried using one of the more successful generations as the basis for another image - Cindy's first therapy session with Crystal, and how I described her in that. The AI really struggled with this one - first, simply making the babydoll into something a bit more leather-like - and really resisting the idea of a mesh top (or any kind of top!) over the bra. Weird. It struggled with colours as well. When I tried to add a bit of pink to lips and fingernails, it turned her hair and clothes pink as well. The AI, at least in this drawing model, also struggled with the idea of an "angry look". Still, these complaints seem pretty petty considering what an absolute numpty like me can create with these things! AI is exciting and terrifying in equal measure....
In any case, I'll leave the AI Art for the time being (but welcome contributions, AI or otherwise) - I've got writing to do!
David Sanders
2023-12-30 17:02:45 +0000 UTCAsklepios
2023-12-30 12:38:41 +0000 UTC