Lucrecia from Counter:Side 카운터사이드

This model was trained on a single game cg and several AI generated images.

Note:
This is a B-LoRA. Use the style LoRA to adjust the effect of game cg style.

Prompts:-

• Trigger Words:

lucrecia_olis, 1girl, purple hair, hair over one eye, blue eyes

• Main Outfit:

neck ribbon, cleavage, fur-trimmed cloak, see-through leotard, asymmetrical legwear

Optional:
circlet, mouth veil
single fingerless glove

• Optional:

expressionless

• Negative Prompts (optional):

circlet, mouth veil, single fingerless glove, staff, holding staff
simple background

Weights Tested:

Character LoRA: ~0.8-1
Style LoRA:     ~0.5+ for a more accurate outfit
                ~1    for a more accurate game CG style

(Feel free to experiment with the weights.)

Known Issues:

• The model may be a bit overtrained, as I prefer a more accurate outfit. Feel free to reduce the LoRA weights if needed.

• You might need to increase the weight of certain hand poses, since the game CG has only one pose.
  

Afterword:
This is my first time working with a single image. I started by cropping it into various proportions and then used these to train a model. Afterwards, I generated some images with this model and cherrypicked those with different outfits and angles to reduce outfit bleeding. Finally, I trained a model with both these AI images (using a lower loss weight) and the original images.

Hopefully some of you will like this model.