Another thought: Your post’s conclusion suggests that good use of the models under discussion is about the quality of the end result. I disagree.
There are far too many flaws in the processes of using the models like Tom and maybe others around him have been.
Even if a comparison of the latest Ren’Py versions with Ren’Py from five years ago wasn’t showing a gap in understanding of what the program is supposed to be doing vs. what it actually does, the past five years are full of missed opportunities, wasted resources, and actual harm that are, in my mind, relevant.
I’d prefer a sloppy painting by a pig be added to project than to any AI-generated image that could fool that public as coming from an acclaimed artist. The appearance doesn’t matter weighed everything else.