OpenAI released their 4o edition of image gen a few days ago, prompting the usual discussion on xitter and elsewhere. The thing is, I still have yet to see a solid video game or comic or similar mostly featuring AI art. I guess it still has trouble with keeping things consistent and whatnot, although 4o specifically seems to have made steps towards addressing this. Regardless I've seen a fair bit of pretty appealing but one-off, non-sequential pictures made by AI (on pinterest mostly) so I feel like it shouldn't be that far out of the question--given a reasonably competent user--but apparently not. On the subject, my own experiments have always ended up going around in circles (I tried with 4o while writing this and had the usual result, although it definitely did stay "on model" better [we will ignore the fact that said model was not really what I had in mind]) as I try to coax the AI down my envisioned path. I guess maybe it just isn't possible, although one other factor is that I think AI tends to have better results when you don't have something very specific in mind, which for the purposes of our investigation here is quite unhelpful. It's not that I've changed my mind about anything I wrote last time, and I think the reasoning is solid but even so it's still hard to believe. Curious..