v1.1a:
v1.1的微调版本，可以通过添加不同的trigger更好的控制视角
共通的trigger为1girl,1boy,sex
配合from front可以生成正对女性的视角
from side同理
sex from behind,pov可以生成pov视角
from normal可以生成常规的视角
评论区提到的??效果我没特意训练，但是试过?????? grab, grabbing ???????????,可以实现
The following is machine translation
v1.1a:
The fine-tuned version of v1.1 can better control the viewing angle by adding different trigger.
The common trigger are 1girl, 1boy and sex.
Cooperate “from front” to generate a perspective facing women.
“From side" in the same way
"Sex from behind,pov" can generate pov perspective.
"From normal" can generate a regular perspective.
I didn't specially train the effect of grasping the ????? mentioned in the comment section, but I tried "?????? grab, grabbing ???????????," which can be achieved.
================================================================================================
v1.1:
根据评论区老哥的意见，剪除了一些tag重新训练了模型，同时还增强了不同视角下的效果，添加了触发词
权重在0.5到0.7之间，同样感觉没什么影响
试了anything 4.5，parstelmix，counterfeit2.5，smilex，perfectworld，chilloutmix都能正常使用，其中anything4.5效果比较好，conterfeit2.5可能要适当调高权重
通过from side和pov能够得到相应的效果，???????????效果和pov类似，但是经常生成女上位背对的图片（???????????中女性更高概率会回头，而pov中不会，我不清楚其中原理）
虽然并非我本意，v1.1生成有具体脸的男性概率增大了，而且大多都是身材健美的清爽帅哥):
可以和其他lora一起使用，我尝试了几个角色lora效果都还不错
The following is machine translation
v1.1:
According to the comments of the elder brother in the comment area, some tags were cut out to retrain the model, and at the same time, the effects from different perspectives were enhanced and trigger words were added.
The weight is between 0.5 and 0.7, and it doesn't matter.
Anything 4.5, parstelmix, counterfeit2.5, smilex, perfectworld and chilloutmix are all used normally, of which anything4.5 works better, and conterfeit2.5 may need to increase the weight appropriately. Corresponding effects can be obtained through from side and pov. The ??????????? effect is similar to pov, but it often produces pictures of women facing back to back (women in ??????????? are more likely to turn back, but not in pov, and I don't know the principle).
Although it was not my intention, the probability of v1.1 producing men with specific faces has increased, and most of them are handsome and refreshing handsome men):
It can be used with other loras. I have tried several roles' Lora,it works well.
================================================================================================
v1.0:
不知道英文怎么说(:
用anything4.5练的，试了几个模型都能用（包括真人），总的来说感觉anything4.5效果最好
权重选在0.5到0.7，好像影响不是很大。
如果不添加修饰词，偏向于生成蓝色系和白色系的头发。默认的视角就是从侧前方，如果添加from side，可以变成从侧边；pov似乎是无效的；??????????? 会生成正确视角的女性，但是男性会出现在奇怪的位置。
即使不添加男性或者???的修饰词，也会默认生成和男性的???。
可以和其他lora一起使用。
The following is machine translation
v1.0
I practiced with Anything 4.5, and tried several models (including real people). Generally speaking, I feel that Anything 4.5 has the best effect.
The weight is chosen from 0.5 to 0.7, which seems to have little impact.
If you don't add modifiers, you tend to generate blue and white hair. The default angle of view is from the front of the side. If "from side" is added, it can be changed to from the side.Pov seems to be invalid; ??????????? will produce women with the right perspective, but men will appear in strange positions.
Even if you don't add the modifier of male or ?????? behavior, you will generate ?????? behavior with men by default.
Can be used with other lora.
