New Release: <a target="_blank" rel="ugc" href="https://civarchive.com/models/21134/or010">林忆宁，古户绘梨花 (010, Furuki Edoka)</a><h3>X/Y Tests:</h3>Prompt test:<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/961d59a2-5412-4470-4935-1b6dca67c800/width=525/961d59a2-5412-4470-4935-1b6dca67c800" />CFG test (w/ 010_v32-000016 model):<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8869045b-faf1-41df-9be1-3e62def77700/width=525/8869045b-faf1-41df-9be1-3e62def77700" />Generalization test (w/ 010_v32-000016 model):<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9e241e34-90da-4194-6788-db365730fa00/width=525/9e241e34-90da-4194-6788-db365730fa00" /><h3>Dataset:</h3>Images chosen from 010's selfie set, specifically during her days as an acitve member of SNH48's Team X. Training:This LoRA model is trained with <a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss">Kohya_ss</a> GUI. Please check version descriptions for training parameters <h3>Tips:</h3><ol><li>I suggest using this model at 0.6~0.9 weight for best results. Going above 0.9 usually causes distortion in face.</li><li>Low CFG scale around 2.5 to 5.0 is recommended. Turn it down if you are seeing oversaturated images.</li><li>Using this model to generate NSFW content is possible, but highly dissuaded.</li></ol><h3>Reach Out:</h3>七海电报群：<a target="_blank" rel="ugc" href="https://t.me/nanamivr">https://t.me/nanamivr</a>宁宁肉体厨：<a target="_blank" rel="ugc" href="https://t.me/nanami_ai">https://t.me/nanami_ai</a>Any questions, suggestions and ideas on improving the model are appreciated.

林忆宁，塞纳河小偶像 (010, idol girl from SNH48)

<ul><li>Minor fixes on the captions on the dataset images</li><li>Clip size changed to 1, from 2</li></ul>This should be the finalized version of this LoRA model, unless I discovery some major flaws in the training process...

Files with SHA256: 05ebf6c2577ff02e50ada0e3654e58f6f989341faab403255fc78f6401c83a8a