First model I've tried training, obviously, being that the character is an original character of the artist Aka6, the data set was small as there are only 6 images of this character and all of which have the same art style, so the model will probably have an effect on the style of your generation.

I have however tested it with other art style LoRAs, and it does work.

Best results I got were using strength between 0.5-0.65 with the Aka6 style LoRA at 0.9, of course including all the keywords, as well as the PerfectEyes LoRA and some negative embeddings.

Aka6 Style LoRA: https://civarchive.com/models/336244?modelVersionId=376556

PerfectEyes LoRA: https://civarchive.com/models/118427?modelVersionId=128461

Negative embeddings: https://civarchive.com/models/831971?modelVersionId=930629 , https://civarchive.com/models/332646?modelVersionId=720175