专为中文用户写的声明：别来我这发乱七八糟的推广小广告，不感兴趣、不会回复、会直接拉黑。国内互联网随地大小便的陋习别tm带到这来。

What is this

LoRA版本的极乐迪斯科风格模型, 使用共计15张图片, 分辨率为1024*1024, 用Anything v4.5训练. 其他模型没有经过测试, 你可以自行尝试效果.

This is the LoRA version of art style similar to the famous and award-winning video game "Disco Elysium".

Trained with only 8 images for portrait and 7 images for scenery, under 1024*1024 resolution. The result is satisfying so I decide to release it.

It was trained on anything v4.5. Other 2d anime models should be fine too, feel free to try it and see for yourself.

Recommend to turn on Hires fix otherwise image lacks of details.

❓How to use / 使用方法

可用于生成角色肖像及风景.

This LoRA is capable of generating both character portraits and scenery.

For portrait👨, the trigger words are / 肖像提示词:

🔸ZaUm, elysiumChar, portrait, <DiscoElysiumStyle_v1:1>,

For scenery🏞️, add these / 风景提示词:

🔸ZaUm, elysiumScape, scenery, <DiscoElysiumStyle_v1:1>,

📝其他信息与参数:

Model: Anything v4.5,

Image Resolution / 分辨率: any, tends to work better on high resolution

Sampler / 采样器: DPM++ SDE Karras,

Sampling Steps / 采样步数: 20-40,

Hires fix / 高分辨率修复算法: R-ESRGAN 4+.

💛 I hope you enjoy it as much as I do. 希望你喜欢这个模型.💛