AI绘画 dreambooth模型训练尝试——以丽芙为例

首先，先展示几张图看看模型最终的效果。

parameters

masterpiece, best quality, best 8k wallpaper, novel illustration,1 girl, solo, bangs, white hair, grey eyes, sidelocks, medium breasts, updo,1 girl wears uniform and belt and long skirt and gloves and thigh boots with armband and necktie, laurels,night sky, full moon, stars, galaxy,

Negative prompt: {Multiple people},lowres,bad anatomy,bad hands, text, error, missing fingers,extra digit, fewer digits, cropped, worstquality, low quality, normal quality,jpegartifacts,signature, watermark, username,blurry,bad feet,cropped,poorly drawn hands,poorly drawn face,mutation,deformed,worst quality,low quality,normal quality,jpeg artifacts,signature,watermark,extra fingers,fewer digits,extra limbs,extra arms,extra legs,malformed limbs,fused fingers,too many fingers,long neck,cross-eyed,mutated hands,polar lowres,bad body,bad proportions,gross proportions,text,error,missing fingers,missing arms,missing legs,extra digit,
Steps: 80, Sampler: DPM++ 2S a Karras, CFG scale: 11, Seed: 2886423289, Size: 512x728, Model hash: 7746fc20, Batch size: 4, Batch pos: 0, ENSD: 31337

加上翅膀和光环就更像了。

parameters

masterpiece, best quality, best 8k wallpaper, novel illustration,1 girl, solo, bangs, white hair, grey eyes, sidelocks, medium breasts, updo,1 girl wears uniform and belt and long skirt and gloves and thigh boots with armband and necktie, laurels, mechanical wings, mechanical halo,night sky, full moon, stars, galaxy,

Negative prompt: {Multiple people},lowres,bad anatomy,bad hands, text, error, missing fingers,extra digit, fewer digits, cropped, worstquality, low quality, normal quality,jpegartifacts,signature, watermark, username,blurry,bad feet,cropped,poorly drawn hands,poorly drawn face,mutation,deformed,worst quality,low quality,normal quality,jpeg artifacts,signature,watermark,extra fingers,fewer digits,extra limbs,extra arms,extra legs,malformed limbs,fused fingers,too many fingers,long neck,cross-eyed,mutated hands,polar lowres,bad body,bad proportions,gross proportions,text,error,missing fingers,missing arms,missing legs,extra digit,

Steps: 80, Sampler: DPM++ 2S a Karras, CFG scale: 11, Seed: 1845882932, Size: 512x728, Model hash: 7746fc20, Batch size: 4, Batch pos: 1, ENSD: 31337

我觉得最终效果相比之前embedding和hypernetwork模型比起来好很多了（相同训练集）。

当然，这些模型也可以一起叠加使用，加强效果。

叠加使用了之后经典粉红色的逆元装置也长出来了，因为训练集里也有极昼的数据所以也可能长不出来，眼睛颜色也是粉色和灰色需要其他词缀配合调整，如果一定要指定的话那就要在训练数据里只有一种装甲的样子了。

总结一下就是dreambooth对于人物识别效果更好，以后可以试试用更少的特定数据训练，看看能不能再提升效果。有兴趣的指挥官也可以自己试试训练其他人物或者战双风格的模型，我会把教程链接放在下面。

参考教程：

PS：租个3090一小时就1.5元左右还有代金券可以用，一个模型大概20分钟完事，大家尽量在云端训练比较好。

模型需要的话可以下载：

链接：https://pan.quark.cn/s/a4064231f8e9