明日から本気出すLoRA model for Hatsuyuki from Kantai Collection.Trained using 190 images. Many thanks to all the artists. <s>k2 where</s><h1 id="heading-1822">Recommended settings</h1><ul><li>LoRA network weight around 0.8-1.0</li><li>Please generate using highres fix for best effect.</li><li>For eye expression it can help to use specific lora such as: <a target="_blank" rel="ugc" href="https://huggingface.co/2vXpSwA7/iroiro-lora/blob/main/test/eye-hanme_01.safetensors">hanme1</a> by2vXpSwA7, <a target="_blank" rel="ugc" href="https://civarchive.com/models/93681/eyecontrol-jitome">jitome</a> by ManaMomo, or other similar</li></ul>For LoRA Block Weights:<ul><li>OUTD (1,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) seems pretty safe</li></ul><h1 id="heading-1823">Suggested prompts</h1>Base appearance (no rigging)<pre><code>1girl, hatsuyuki \(kancolle\), 
blunt bangs, hime cut, black hair, long hair,
serafuku, teal sailor collar, short sleeves,
pleated skirt, blue skirt,
black socks, anchor symbol,</code></pre>There's only one main outfit baked in, so you can usually get away with less tags than this. Also you can try "(blue) sailor collar" but it might not turn very blue.Rigging<pre><code>smokestack,
machinery,
rigging,
torpedo launcher,
thigh strap,
backpack,
holding turret,
cannon,
armored boots,</code></pre>Rigging is quite gacha. Try controlnet or other methods to at least limit the search space.You might need to add some tags if your base model draws her a bit too big (flat chest, young, etc.)

Files with SHA256: 55b802cdcea9b63c994b464c20d6b55948ae4fb1c7ba6a2284cb69227c4dfb45