This is derived from four models on C Station. I remember the original models were GeminixMixRealistic and cndasianmix_av. Compared to Fancha 08, the colors should be a bit richer, with slightly more details in the background. The blended facial shapes also have a bit less AI feel. Feel free to use if you like. One more choice.

The cover image is a live image from lib online.

I suggest creating more local live images. It's a bit strange that with the same parameters for local and online, using the same CPU random seed source, you still can't generate the same images.

The online effects are really not commendable. Take a look at a few local examples below. Using the same parameters as the cover image, I still prefer the local effects. They are softer.

A small suggestion, when creating half-length portrait images locally, try to avoid public places. This guy will occasionally mischievously give you some shots with revealing points,...

Oh, by the way, someone asked before if each of my models is face-locked. Answer: No. None of the models I blend have any added features or face types like lora.