very nice!

Good to see her. She'd been on my "I should do this" list for a while, but 1: getting all the angles needed for 3D model based models takes a decent amount of effort since I'd need to reinstall the game and preferably keep the backgrounds and poses different 2: I was worried how her segmented floated hair would be understood. 3: decide which costumes to include
A few tips: "average proportions" (and indeed, most tags about build) is useless at best or counter purpose at worst. Your image selection is teaching the proportions. You only want to tag proportions when it deviates from your standard in an image (and you only want to include such images if you're really desperate for images, or you want multiple variations of a character in one LoRA)
Color should come first in tagging, not middle. You don't even really need to specify color in tags (it will default to what it was trained on) unless you've got black and white images in the dataset (and maybe for maybe hair color). You also don't need to describe articles of clothing ("large" or how the shoes are designed) beyond their base type. The reason to use specific tags that Illustrious already knows is compatibility reasons more than anything (a style LoRA may have "shoes" hundreds of times and try to change a character's "shoes" to match that baseline, but "high tops" won't be).
I don't know if you did, but I'll point it out since it's not obvious: You can use 3D model screenshots if just tag the image "3D" in training and (preferably) either . That's 12+ pictures when combined with camera tags ("straight-on", "from side", "from behind", "from below", "from above" etc.)