Type | |
Stats | 974 0 |
Reviews | (93) |
Published | Mar 4, 2024 |
Base Model | |
Trigger Words | march_7th_(honkai:_star_rail) march_7th_(princess)_(honkai:_star_rail) march_7th_(kfc)_(honkai:_star_rail) march_7th_(feast)_(honkai:_star_rail) |
Hash | AutoV2 C5A06A17E8 |
3.0版本更新 3.0 version update
目前一共支持三套服装
like：
march_7th_(honkai:_star_rail)
march_7th_(princess)_(honkai:_star_rail)
march_7th_(kfc)_(honkai:_star_rail) （建议权重0.8）
march_7th_(feast)_(honkai:_star_rail)
-
Training model: Animagine XL V3
Inference model: Animagine XL V3
Training resolution: 1024*1024
Recommended output resolution: 1024*1280
Recommended model weight: 0.8~1
2.0版本更新：
终于可以画好眼睛啦！现在可以还原三月七的眼睛，只需要一点点技巧...
1.5更新：
学会了角色三月七的彩色眼睛，但是代价是300张图片，每一张图片训练了1000遍，是的1000遍。
SDXL真神奇，居然还没有过拟合。
之前有可能是负面tag影响了，因此，我推荐：
正面必备tag：sanyueqi,1girl,solo,pink hair,long sleeves,blue eyes,bangs,shirt,white shirt,hair between eyes,pink eyes
负面推荐tag：photo,deformed,realism,disfigured,lowres,bad anatomy,bad hands,text,error,missing fingers,extra digit,fewer digits,cropped,worst quality,low quality,signature,watermark,blurry,
如果眼睛还是不行，下面推荐栏有朋友单独炼了一个眼睛的模型，可以试试/
Update 1.5:
Learned to generate colored eyes for the character "Sanyueqi," but it came at the cost of using 300 images, each trained 1000 times. Yes, 1000 times.
SDXL is truly amazing since there hasn't been overfitting so far.
Previously, negative tags might have had an impact. Therefore, I recommend the following:
Recommended positive tags:
sanyueqi, 1girl, solo, pink hair, long sleeves, blue eyes, bangs, shirt, white shirt, hair between eyes, pink eyes
Recommended negative tags:
photo, deformed, realism, disfigured, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, signature, watermark, blurry
If the eyes still don't look good, there's a separate eye model recommended in the section below, which a friend has specifically trained. You can give it a try
The English portion is translated by GPT and may contain errors.
值得注意的是，在该模型的训练中，我使用waifuc来通过搜集和处理训练集，他可以轻易过滤掉我所不需要的双人图，杂图和黑白图片，这使我的训练集的质量得到提高。
https://deepghs.github.io/waifuc/main/tutorials-CN/installation/index.html
It is worth noting that in the training of this model, I used waifuc to collect and process the training dataset. It can easily filter out unwanted images such as double exposure, miscellaneous images, and black and white images, which improves the quality of my training dataset.