The bad eyes, hands, and inconsistent art styles have mostly been resolved with newer AI models. It mostly boils down to having a good workflow set up and having good source models. There are lots of tools out there now that do things like fixing and upsacaling eyes and hands after you get a good initial image.
The two results in my example is an initial generation and then running that through a tool that masks and upscales the face.
You must be registered to see attachments
*NOTE*
This game was up when I initially wrote this post and made this image to show the current state of AI, then it was taken down before I clicked submit, and now I am seeing that it is up again and my preview of my post is still here. So I am now actually submitting this post. It was initially a response to some comments about AI inconsistencies and bad eyes/hands.