对天涯明月刀OL游戏内人物外观进行LORA模型训练后的成果,v0.2版本采用了全新的参数微调,通过LoRA分层移除了对脸部的训练,使其脸部的生成更贴近大模型的效果,减少了崩脸的概率,但依旧推荐你采用面部修复,v0.2版本的预览图所用大模型为Guofeng3_v34
对模型测试以及炼丹感兴趣的,欢迎来Q群850498709(一个人又训练又测试,太疲乏了而且容易忽略问题)
如果你喜欢本模型,也欢迎你来赞助我,这将是我更新模型的最大动力,赞助链接:https://fpoh.usells.com/p/d9Uu7h
以下是展示预览图的生成信息(C站自动识别不了,一个一个填又太麻烦)
提示词
<lora:[衣服]-[林中鹿-天刀少女]TDSVlzl_v2:0.7>,1girl,(loli:1.2),solo,hairball earmuffs,fur-trimmed_gloves,vertical-striped_legwear,black_thighhighs,white fluffy slippers,happy,standing,(full body:1.6),indoor,tile_floor,christmas_tree,(looking_at_viewer:1.2),(best quality),((masterpiece)),
负面提示词
(ear:1.3),(fingering_through_clothes:1.3),(toes:1.3),(animal toes:1.3),(worst quality:2),(low quality:2),(normal quality:2),
(unclear eyes:1.331),bad hands,missing fingers,bad body,(more than 2 nipples:1.331),(extra legs:1.331),(unclear eyes:1.331),
其他参数
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 2318987238, Size: 680x904, Model: [3D]-[国风]3Guofeng3_v34, VAE: 840K-EMA_VAE.safetensors, Denoising strength: 0.5, Clip skip: 2, ADetailer model: mediapipe_face_full, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 23.9.0, Hires upscale: 2, Hires steps: 15, Hires upscaler: 4x-UltraSharp, Lora hashes: "[衣服]-[林中鹿-天刀少女]TDSVlzl_v2: 562a87e3a5d7", Dynamic thresholding enabled: True, Mimic scale: 7, Separate Feature Channels: True, Scaling Startpoint: MEAN, Variability Measure: AD, Interpolate Phi: 1, Threshold percentile: 100, Version: v1.6.0