In version 2.1, I reduced the number of outfits; the dataset was too large, and I wanted to focus on the swimsuit skin and improve eye detail.
Also, due to the large number of images featuring a swimsuit, the AI learned her athletic physique beneath the swimwear and reduced her bust size; you need to adjust such changes yourself.
Occasionally, the red mark on the neck might appear in an unwanted spot, or two gloves might show up instead of one, but these are minor issues. If you copy the prompt and add it to the "person" detailer, you will get the results shown in the swimsuit skin images.
Description
The dataset contains about 300 screenshots, and there's not a single one where the character is standing far away in a full body. However, for reasons unknown to me, if you don't specify a close-up view, the full set of clothing produces shots with a distant view. However, having a full set of prompts for each skin doesn't necessarily mean you have to use every block of text (for example, don't write about shoes). If you're not generating faces very close up, just use the "red eyes" prompt for the eyes, and then specify closer ones in the detailer: faceted red eyes, crystalline pupils, sharp eyeliner. The eyes are far from perfect, and I've started working on Lora from scratch many times. I couldn't achieve anything more than a satisfactory result for publication, and I no longer have the energy or knowledge to improve the work.
Regarding NTE, I'm focused on fairly unusual characters and won't make just anyone, as I play the game and don't get all the event characters (even though I have two accounts). In the near future I will improve the existing lora alphard (and publish it) and try to work with the old lora zanku who had big problems with her eyes.