这个模型不需要触发词This model doesn't need a trigger word【推荐出图参数】分辨率: 横屏 原生2K 2160x926 2449x816 （21:9 丶3:1 丶16:9）lora权重：1.0采样er_sde + simple ，采样次数10，温馨提醒：krea2的提示词和lora效果本身一样重要，好的提示词能达到更高的上限，建议尝试倒推高质量的图片进行尝试使用了240张黑神话悟空的风景图片进行训练。首次尝试UE5风格的Lora训练，很显然这不是一个成功的Lora, 但是 Krea-2 让我看到了可能性。受限于租卡成本和技术，这个Lora我并不满意，后续我会持续改善，也希望有想要尝试训练的小伙伴联系我，我收集了1000+张训练集，训练恰当的话，他可能达到数倍好于我的效果In Krea2, the prompt is just as important as the LoRA effect itself. A good prompt can raise the ceiling significantly. It is recommended to try reverse-engineering high-quality images for better results.It was trained on 240 landscape images from Black Myth: Wukong.This is my first attempt at UE5-style LoRA training. Obviously, this is not a successful LoRA, but Krea2 showed me what’s possible.Limited by GPU rental costs and technical skills, I’m not satisfied with this LoRA. I will continue to improve it, and I also hope that anyone who wants to try training will contact me. I have collected over 1,000 training images. If trained properly, the model could achieve results several times better than mine.

Files with SHA256: cf66825b87cd537771f9ab501791320b64dffa446e76a4fb977fc96d8fb6d9b6