<h3 id="heading-252">Beware!: These two version(4.0 and 4.1) doesn't really match my goal. I'll retrain her when I have more HQ dataset. Sorry for the bad quality.</h3>(4.0 and 4.1 looks similar but 4.0 maybe has more eye makeup or smth, pick any version you like)Hello it's OKGC, kpop girls/actress lora maker. I speak Mandarin, English, and a little Korean. Follow the info below to get better result. Feel free to ask for help.<h3 id="heading-6554">Recommended checkpoints:</h3><ul><li>supashy mix - <a target="_blank" rel="ugc" href="https://civarchive.com/models/156202/supashymix">https://civarchive.com/models/156202/supashymix</a><ul><li>final version works best (what I'm using), get it on supashy's gumroad to support him please - <a target="_blank" rel="ugc" href="https://supashy.gumroad.com/l/supashy_mix_final?layout=profile">https://supashy.gumroad.com/l/supashy_mix_final?layout=profile</a></li></ul></li><li>Moonfilm Reality 3.0 - <a target="_blank" rel="ugc" href="https://civarchive.com/models/43977?modelVersionId=73071">https://civarchive.com/models/43977?modelVersionId=73071</a></li><li>majicMIX realistic v7 - <a target="_blank" rel="ugc" href="https://civarchive.com/models/43331?modelVersionId=176425">https://civarchive.com/models/43331?modelVersionId=176425</a></li></ul><h3 id="heading-6555">Recommended parameters: (works best in my opinion)</h3><ul><li>Trigger word: kangseulgi</li><li>Sampling Method: Euler</li><li>Steps: 50</li><li>Hires. fix Upscaler: 4x_NMKD-Superscale-SP_178000_G - <a target="_blank" rel="ugc" href="https://huggingface.co/gemasai/4x_NMKD-Superscale-SP_178000_G/tree/main">https://huggingface.co/gemasai/4x_NMKD-Superscale-SP_178000_G/tree/main</a><ul><li>Put it in folder models/ESRGAN</li></ul></li><li>Hires upscale: 2</li><li>Hires steps: 15</li><li>Denoising Strength (Hires): 0.25</li><li>Clip skip: 2</li><li>CFG scale: 7</li><li>Size: 512x768</li><li>vae: vae-ft-mse-840000-ema-pruned - <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt">https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt</a><ul><li>Put it in folder models/Vae</li></ul></li><li>Eta noise seed delta (ENSD: 31337)<ul><li>Find it in SD webui/Settings<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/85372f95-2964-48e3-a613-e3fe2a2d5c44/width=525/85372f95-2964-48e3-a613-e3fe2a2d5c44.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c39729e5-fe17-4bba-b102-54b99d2f99af/width=525/c39729e5-fe17-4bba-b102-54b99d2f99af.jpeg" /></li></ul></li></ul><h3 id="heading-6556">If the face comes out badly, use adetailer to fix it.</h3><a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">https://github.com/Bing-su/adetailer</a>Recommended parameters: (works best in my opinion)<ol><li>1st Adetailer model: face_yolov8n.ptInpaint denoising strength: 0.25<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ac9055c0-c572-41a8-aa31-a289668ac8c1/width=525/ac9055c0-c572-41a8-aa31-a289668ac8c1.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a928fec2-62a3-4ae7-a424-758c7fd4a3e8/width=525/a928fec2-62a3-4ae7-a424-758c7fd4a3e8.jpeg" /></li><li>2nd Adetailer model: mediapipe_face_mesh_eyes_onlyInpaint denoising strength: 0.25<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ac9055c0-c572-41a8-aa31-a289668ac8c1/width=525/ac9055c0-c572-41a8-aa31-a289668ac8c1.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a928fec2-62a3-4ae7-a424-758c7fd4a3e8/width=525/a928fec2-62a3-4ae7-a424-758c7fd4a3e8.jpeg" /></li><li>If the face detection of adetailer detects the face of other people in the background,set Detection model confidence thereshold: 0.8~0.85<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/71a736d4-ca35-46c7-bff0-230592e844a4/width=525/71a736d4-ca35-46c7-bff0-230592e844a4.jpeg" /></li></ol><h3 id="heading-6557">Used embeddings:</h3>(how to use?: <a target="_blank" rel="ugc" href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Textual-Inversion">https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Textual-Inversion</a>)<ul><li>ng_deepnegative_v1_75t - <a target="_blank" rel="ugc" href="https://civarchive.com/models/4629/deep-negative-v1x">https://civarchive.com/models/4629/deep-negative-v1x</a></li><li>badhandv4 - <a target="_blank" rel="ugc" href="https://civarchive.com/models/16993?modelVersionId=20068">https://civarchive.com/models/16993?modelVersionId=20068</a> (add disfigured in negative prompt may also helps getting better hand)</li><li>bad_pictures - <a target="_blank" rel="ugc" href="https://civarchive.com/models/23178?modelVersionId=66043">https://civarchive.com/models/23178?modelVersionId=66043</a></li></ul><h3 id="heading-6558">Highly recommended tools: Lora Block Weight - <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-lora-block-weight">https://github.com/hako-mikan/sd-webui-lora-block-weight</a></h3><ul><li>How to use?&lt;lora:lora name:1:lbw=IN02&gt;  &lt;lora:lora name:1:lbw=1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0&gt;</li><li>Recommended parameters (name whatever trigger word you want, it's all about the block weight)<ul><li>Body:WINGS:1,0,0,0,0,1,0,1,1,1,0,0,0,0,0,0,0BODY:1,1,1,1,1,1,1,1,0,0,0,1,1,1,1,1,1BODY0.5:1,1,1,1,1,1,0.2,1,0.2,0,0,0.8,1,1,1,1,1</li><li>Face: (face shape, hair style, eye shape, eye colors, etc.)EYE:1,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0MOUTH:1,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0FACE:1,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0 (which I use for my lora)FACE0.5:1,0,0,0,0,0,0,0,0.8,1,1,0.2,0,0,0,0,0FACE0.2:1,0,0,0,0,0,0,0,0.2,0.6,0.8,0.2,0,0,0,0,0</li><li>Hands:HAND:1,0,1,1,0.2,0,0,0,0,0,0,0,0,0,0,0,0</li><li>outfit: (with tags)CLOTHING:1,1,1,1,1,0,0.2,0,0.8,1,1,0.2,0,0,0,0,0</li><li>pose: (with tags)POSE:1,0,0,0,0,0,0.2,1,1,1,0,0,0,0,0,0,0</li><li>palette/color style: (with tags)PALETTE:1,0,0,0,0,0,0,0,0,0,0,0.8,1,1,1,1,1</li><li>Character:KEEPCHAR:1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0</li><li>Background:KEEPBG:1,1,1,1,1,1,0.2,1,0.2,0,0,0.8,1,1,1,0,0</li><li>Reduce overfitting (same as OUTALL）REDUCEFIT:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1</li></ul></li></ul>

Seulgi

The output looks similar to 4.1, pick any version you like

Files with SHA256: bdd3d48b5dedca086d0fba84ed93a5f648361673cd56a35f3b8a9a394807b287