头发的缺陷看起来像pony，有明显的过拟合的痕迹。看画风，盲猜一个t-pony nai3，是一款用pony模仿nai3的画风的模型。

本区一般认不出来画风的
这种有点乱的头发``````米山舞（画师名）？
用画师名试试……不太像
下图是某个米山舞LORA的例图

发现有点像quasarcake

找了一年半？？？没必要，花钱快速解决。。。。

上civitai找类似的看底模和参数就行

眼睛一眼加了gomzi

是想要画风么...找了一年半学学炼丹搜搜素材都够炼几个画风丹出来了

竟然在首页刷到我去年发的图了，这个图是我用novelai3生成的，已经过去好久了，我那个原p站号被封了，谢谢你的喜欢。

这是叠加了氛围转移的图，tag词：正面提示词：dynamic angle,{((elysia_(honkai_impact))},((pink_hair)),{artist:shiina kuro},(artist:ask_askzy),{artist:ningen_mame},[[airtist:atdan]],{{artist_ciloranko}},{artist:wlop}, [artist:yoneyama_mai],artist:ogipote,artist:morikura_en,noyu(noyu23386566),artist:morikura_en,year_2023,konya_karasue,
shadow,colorful,blurry foreground,depth of field,
1girl,long_hair,tears,nude,hairflower,colorful butterflies,angel wings,standing,head up,jirachi,flowers,singing,wind,stars,holy,junketsu,lighting,shards,Mew,close up
,best quality, amazing quality, very aesthetic, absurdres,highres
负面：nsfw, lowres, {bad}, error, fewer, extra, missing, worst quality, jpeg artifacts, bad quality, watermark, unfinished, displeasing, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract], lowres, {bad}, error, fewer, extra, missing, worst quality, jpeg artifacts, bad quality, watermark, unfinished, displeasing, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract], {{normal quality}}, {{bad quality}}, {{low quality}}, {{worst quality}}, {{lowres}}, displeasing, very displeasing, {bad}, {{poor eyes}}, {{average eyes}}, {{chaotic eyes}}, no pupils, bad anatomy, {{bad hands}}, text, error, missing, {{missing finger}},{{multi-fingered}}, extra, extra digits, fewer, fewer digits, cropped, JPEG artifacts, signature, watermark, username, blurry, artist name, fat, duplicate, mutation, deformed, disfigured, extra arms, extra legs, long neck, {{bad feet}}, sagging breasts proportions, extra, fewer, unfinished
参数：
steps:28
height:1216
width:832
scale:7
uncond_scale:0
cfg_rescale:0.5
seed:1525663850
n_samples:1
hide_debug_overlay:false
noise_schedule:"native"
legacy_v3_extend:false
reference_information_extracted_multiple:[
0:1
1:0.2
2:0.2
]
reference_strength_multiple:[
0:0.15
1:0.15
2:0.15
]
sampler:"

直接找相同画风的图的模型

①找原图
②找类似图

好看

建议直接问原作者