
该lora在nai1上进行训练
训练集主要由Danbooru、网页截图和二创作品(ai画)组成,包含少数的游戏截图
对身体各个部位做了加强训练,可以通过相应的tag唤出全身,上半身,脸部特写特征等
无法较好的还原游戏中的形象,但泛用性高,有一定可玩性,建议搭配cetus-MIX WhaleFall2模型使用,权重推荐0.6~0.8
您也可以通过调整prompt画出流萤的千姿百态
注:训练中加入了评级标签,分为四个等级:safe, sensitive, nsfw, explicit,可按需求自行使用
希望大家多多返图并提出宝贵的意见,后续我也会积极改正
大多数时候无特需请尽量不要使用负面嵌入(会使你的风格固定在某一个特定方向),有兴趣可自行尝试,或许会收获更好的结果
The lokr is trained on nai1
The training set consists mainly of Danbooru, web screenshots, and second creations (ai art), with a few game screenshots
The various parts of the body have been strengthened training, and the whole body, upper body and face close-up features can be called out through the corresponding tag
Can not be better to restore the image in the game, but the universality is high, there is a certain playability, it is recommended to use with cetus-MIX WhaleFall2 model, the weight is recommended 0.6~0.8
You can also adjust the prompt to draw a variety of fireflies
Note: A rating label has been added to the training, which is divided into four levels: safe, sensitive, nsfw, explicit, which can be used on demand
I hope you will return the picture and put forward valuable comments, and I will actively correct it in the future
Most of the time, please try not to use negative embedding (it will make your style fixed in a certain direction), and if you are interested, you can try it yourself, perhaps you will get better results
描述:
训练词语: LiuYing
名称: LiuYing.safetensors
大小 (KB): 26797
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success