This is a set of LoRA for generating the submarine kanmusu from Kantai Collection (Kancolle). Currently, I-168, I-58, I-8, I-19 and I-401 are supported, but I hope to cover them all one day. For now, I'm making individual character models but I want to create combined models eventually as well.
Next sub: I-26
Recommended Weights
Weights around 0.7-1.0 recommended, start at 1 and scale back if it's too strong.
For best effect use with highres fix.
Recommended Block Weights
Use with LoRA Block Weight extension for Automatic1111 WebUI. It's a powerful tool that allows you to tune what parts of the LoRA apply.
Mainly character features
OUTD:1,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0
Character features + stylization
OUTALL:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1
Posing and character features (affects background as well)
INS_MIDD:1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0
Suggested prompts
See "About this version" for each model
Suggested models
Trained on NAI (final-pruned.ckpt [89d59c3dde]) so it should be compatible with most SD1.5 anime models. However I've noticed some have a propensity to turn sukumizu into bikinis so you may need to adjust prompts for those.
I mainly use Anything-V5-prt-RE, it's very versatile and easy to work with.
TIs
You don't really need any of these but they're helpful
negative_hand-neg https://civitai.com/models/56519?modelVersionId=60938
bad-hands-5 https://huggingface.co/yesyeahvh/bad-hands-5/tree/main
verybadimagenegative https://civitai.com/models/11772?modelVersionId=15556
easynegativev2 https://huggingface.co/gsdf/Counterfeit-V3.0/tree/main/embedding
supercha-negv2-p1 supercha-negv2-p2 https://civitai.com/models/56033/supercha-neg
Description
LoRA model for Iku, also known I-19, from Kantai Collection.
Recommended Weights: weights around 0.7-1.0 recommended, start at 1 and scale back if it's too strong.
If you want her nametag, as close to 1 as possible is best.
Suggested prompts
Default outfit (old sukumizu w/ nametag, ribbons, hair accessories)
kciku, one-piece swimsuit, name tag, (19), hair ornament, twintails, multicolored hair, long hair, hair ribbon, large breasts
Things to try
torpedo, torn swimsuit, school swimsuit, old school swimsuit, red eyes, submerged, underwater, freediving, swimming, :d, fang
Symbol Pupils
If the eyes are large enough it can sometimes do star-shaped pupils with emphasis on (((star-shaped pupils))) and ((star \(symbol\))). However it is difficult for VAE to get this level of detail in non-closeups. I recommend using inpaint if you want them.
I tried doing hearts and flower shapes which were also in the dataset but it doesn't seem to work well.
Recommended Block Weights
Use with LoRA Block Weight extension for Automatic1111 WebUI. It's a powerful tool that allows you to tune what parts of the LoRA apply.
Mainly character features
OUTD:1,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0
Character features + stylization
OUTALL:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1
Posing and character features (affects background as well)
INS_MIDD:1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0
Suggested models
Trained on NAI (final-pruned.ckpt [89d59c3dde]) so it should be compatible with most SD1.5 anime models. However I've noticed some have a propensity to turn sukumizu into bikinis so you may need to adjust prompts for those.
I mainly use Anything-V5-prt-RE, it's very versatile and easy to work with.
FAQ
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.