I've been working on ways to get further detail out of my anime images, and this approach is the best I've come across. It gives nice skin texture and facial definition.

It will transform faces on more stylized images. However, if your images are semi-realistic enough it will refine facial details instead of transforming them, which is my favorite application of this workflow.

Flux.2 Klein is getting a bit old now in terms of AI development, but until (if) some LoRAs are made for Qwen 2.1 this still seems to be the best (local) option for this sort of detail refinement.

I see a few workflows out there, but none that had it configured cleanly, so I thought I'd share this for others who are interested (as I plan to use it for some of my images, and can direct people to the workflow here).