大家好，这个lora是一个风格lora，主要是游戏CG，原画类的风格，偏向于拳头游戏旗下的符文之地传说卡牌插图风格，搭配一些写实，原画大模型可能效果会更好，二次元也还行。
注意：可能是训练集的场景图不多，所以场景方面单用这个lora，效果可能要多刷几次。出人物的话，搭配一些游戏人物(尤其是LOL的人物)效果还可以，触发词是 英雄联盟风格，但没有写它，似乎也能出一些效果。
正如我前面说的那样，这不是一个角色lora，这是一个风格lora，正如例图所示，角色和场景都可以，这适用于大部分角色lora(这一版可能效果不是很理想，后续会更新，也会尝试训练XL版本的)，如果可以，欢迎大家返图，也可以提一提意见，以帮助后续我可以更好的改进它，拜托了~~~
请享用
使用这个Lora的推荐设置：
● 出场景时，大模型推荐使用 lyriel、ghostmix、dreamshaper；出人物时，大模型推荐使用revAnimated_v122、ghostmix、dreamshaper，
如果场景和人物都想出二次元类的，建议大模型使用 anything-v5-PrtPE
● VAE建议使用 vae- FT-MSE-84000-EMA-Pruned，Clip skip:2，CFG scale：5~8，lora weight：0.4~0.8
● 采样器：DPM++ SDE Karras、DPM++ 2M Karras、Euler a、DDIM；迭代步数：25~45
● 高分辨率修复(Hires. fix): ESRGAN_4x Latent，放大倍数：1.5~2，重绘幅度：0.3~0.55
警告：
● 您将独自承担因不道德使用此模型而产生的任何法律责任
● 不要用我的模型作为诱饵，在未经允许的情况下驱使人们使用你的付费发电服务
● 除一般个人用途外，请务必与我联系
Hello, everyone, this lora is a style lora, mainly the game CG, the style of the original painting class, tends to the rune land legend card illustration style under the fist game, with some realism, the original painting large model may be better, and the two yuan is also OK.
Note: It may be that there are not many scene diagrams in the training set, so using this lora for the scene side alone may require several more brushes. Out of the characters, with some game characters (especially LOL characters) effect can also be, the trigger word is the league of Legends style, but did not write it, it seems to be able to produce some effects.
As I said before, this is not a role lora, this is a style lora, as shown in the example figure, the role and scene can be, this applies to most of the role lora(this version may not be very ideal, the subsequent update, will also try to train XL version), if you can, welcome to return to the picture, you can also give suggestions, To help follow-up I can better improve it, please ~~~
Please enjoy
Use this recommended setting for Lora:
For large scene, you are advised to use lyriel、ghostmix、dreamshaper. For people, it is recommended to use revAnimated_v122、ghostmix、dreamshaper for large models.
If both scenes and characters come up with quadratic metaclasses, it is recommended to use anything-v5-PrtPE for large models
For VAE, it is recommended that vae- FT-MSE-84000-EMA-Pruned,
Clip skip: 2, CFG scale: 5~8, lora weight: 0.4~0.8
Sampler: DPM++ SDE Karras, DPM++ 2M Karras, Euler a, DDIM; Sampling Steps: 25~45
Hires. fix: ESRGAN_4x Latent, Upscale by: 1.5~2, Denoising strength: 0.3~0.55
Warning:
You are solely responsible for any legal liability resulting from unethical use of this model
Don't use my model as bait to drive people to your paid generation service without permission
For anything other than general personal use, please be sure to contact me