LoRA model for the Hiburi-class from Kantai Collection.

Trained using 329 images. Many thanks to all the artists.

Recommended settings (v9)

• LoRA network weight around 0.6-0.8

• Please generate using highres fix for best effect.

• Effect on quality can be pretty poor since there's not a lot of high quality images for the dataset. I suggest using with detail adding lora or strong aesthetic checkpoints.

• For Shounan's eye expression it can help to use specific lora such as: hanme1 by

  2vXpSwA7, jitome by ManaMomo, or other similar

For LoRA Block Weights:

• No rigging: OUTD (1,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) works well

• With rigging: INS_MIDD (1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0)

Suggested prompts

Common Features

sailor dress, blue sailor collar,
sailor hat,
shoulder bag, shoulder strap,
white socks, shoes,

Hiburi

kchiburi, short ponytail, short hair, brown eyes, black hair,

For some reason she ends up with a thin left sidelock fairly often, please inpaint those out.

Daitou

kcdaitou, bob cut, ponytail, blue eyes, black hair, 

Shounan

kcshounan, short twintails, purple eyes, black hair,

Rigging

Very gacha, I suggest using controlnet to at least set the pose up properly

thigh strap, (depth charge),

machinery, rigging, smokestack, mast,

holding cannon, adapted turret, cannon,

You might need to add some tags if your base model draws them a bit too big (flat chest, child, etc.)