Haven't played the game, just browsed thread, looking at art. Wanted to express my opinion on picture aspect alone. On technical level, I like the idea of enhancing renders by AI pass a lot (I have some knowledge about 3d rendering techniques from both game and DAZ, not so much about post-processing/AI). And results seems very promising. But as seen on the front page of this thread and very nicely shown on the pic quoted above, I find that the end result is missing the mark by overdoing the good stuff. Someone else said as feedback that the girl looks like too much makeup, but I disagree, rather too little or bad one. It is too much detail, too harsh lines, veins and muscle lines and skin detail/age marks, something makeup is usually made to tone/hide. There are pictures here are there with softer light that look quite appealing, but most look too harsh, even if scene suggest softer lighting like on picture above. It's almost cell shaded effect (not quite), where all the details are drawn with black outline to make it imitate pencil work art. So lets just call it oversharpened.
Daz render looks awful in your comparison. Daz can do much better. Expression and lighting are comically bad in every comparison pic I have seen so far (maybe it is enough for your workflow, but still, man wonders what can be done with decent renders). But the kicker is #1 look best for me. The Harshness I was talking about is not yet present in this one, it seems to be added by hand in photoshop in #2. Is making her (and most characters in game as seen on other screens) look so much older and weathered your intention? Clearly, tastes come to play here as well. But in comparison, #1 in already adding all the lovely detailed look to the body and face, with appropriate softens to make it appealing/alluring.
AI/ML seems to me both exciting in what can it do and disappointing how poorly it seems to be serving storytelling because of lack of consistency, and your (and one or two similar projects) approach and technique seems most promising so far.
I have been learning about DAZ for a while. Not an expert, but I have some grasp of possibilities and difficulties. Your technique does seem to offer some existing enhancements. Learning photoshop from scratch (and paying subscription) not in my time/money budget for now, and frankly doesn't seem that exciting. I have more superhero/SF renders in mind and special/particle effects, like energy beams, power shields and explosions is something I haven't seen done well in DAZ (they added VDB support, but libraries are limited) so I'm wondering if to your knowledge is it something that can be done by AI (enhanced I mean, adding simple emissive beam in DAZ is not a problem, it just doesn't look any good, especially compared to all this lovely characters, costumes and locations)? Anyway, mostly giving my 2 cents of feedback, but if you could give me some pointer to where to start learning about AI technique similar to yours, I would be grateful I mean not from text prompt, as most like to be, but from mostly done picture with text prompts to modify it. Should I start with Stable Diffusion (from my googling it is best offline one and free)?