This is my first attempt to train a Lora model, I don't expect much, but any feedback is much appreciated.
This was trained using a few low quality images as there isn't a lot of images of this character going around. At strength 1 the image generated looks close enough to the real thing but sometimes you'll get artefacts on the image generated, as the source material used to train was not that great. At 0.8 you'll still get some spots here and there on the skin of the character in the nude.
About the pokies, it's probably inevitable cosidering the source images were in various stages of nudity. Urushihara Satoshi didn't draw a lot of images of this character and what is available were from his own doujin works.
Considering the quality of the source material, will probably leave this Lora as it is until I find better sources.
You may want to include gloves and skirt to negative prompts as they tend to show up unintentionally
Basic prompts for her:
tricia_growlanserIV, orange hair, bow, hair bow,
Version 2 notes:
This should be better. Can't test it with prefect using on-site generator. But It definitely should be a lot better.
Her updated prompts:
tricia_growlanseriv, orange hair, short hair, hair bow, green eyes,
Her usual outfit:
green dress, collar, short dress, green thighhighs, elbow gloves, earrings,
https://civitai.com/models/1272443/tricia-growlanser-iv
