• 一种“较”真实系二次元风格,在经过一些神秘的lora配比后在k2上获得的风格(大嘘)
• 此风格聚焦于“较”真实的服装细节、皮肤以及良好的背景和光影表现,而不是类似3d的真实,更不是完完全全的真实质感
• 该风格画面较暗
• A "bit" realistic anime style, achieved after some mysterious LoRa matching
• This style focuses on a "bit" realistic clothing details, skin, as well as good background and lighting effects, rather than a 3D-like realism, nor is it a complete realistic texture
• The style of the image is relatively dark
如果不嫌麻烦可以多返点图和评论,让我知道还有哪些不足,感谢🙏
If you don't mind the trouble, you can post more pictures and comments, so I can know what else is lacking. Thank you🙏
Moss base 1.0
• 基于moss base训练,base和sft版本都可用
• 使用降步lora可能会导致部分模糊,而加速节点可能会导致肢体错误,看情况使用
• 该lora可能会导致部分人物lora或者底膜已有的人物特征不明显或失效
• 测试过程中发现提示词尤为重要,较差的提示词和一些有关真实、照片类的提示词可能会导致画面直接真实化,完全没有二次元的画风(尤其是欧美私人卡姿兰多睫毛浓眉大眼!!!),所以注意你的提示词,如果不想出现完全的真人
• 推荐添加anireal和dlss5这两个lora来增强质感和光照,在下面的推荐资源中
• Based on moss base training, both base and sft versions are available
• Using lora with reduced step(4 or 8) count may lead to partial ambiguity, while accelerating nodes may cause limb errors. Use it based on the situation
• This Lora may cause some characters' Lora or existing character traits in the base film to become unclear or ineffective
• During the testing process, it was found that prompt words are particularly important. Poor prompt words and some prompt words related to reality or photos may lead to the direct realism of the image, completely lacking the anime style. Therefore, pay attention to your prompt words if you don't want to see a completely realistic image
• It is recommended to add anireal and dlss5, two loras, to enhance texture and lighting. These are included in the suggested resources below
Moss(sft) v1.0
• sft版本在魔搭平台上训练,按moss模型的说明,sft版本训练出来的lora不适用于base版本 很快啊,一有魔搭能训练的风声就马上开炉😍
• 据魔搭平台的训练反馈,在此平台上训练的lora可能需要更新到最新comfyui版本才能被正确加载
• 后续会尝试训练base版本
• 感觉moss训练出来的lora比k2和anima的效果更显著
• 该版本可能存在部分细节模糊的问题
• The sft version is trained on the modelscope. According to the instructions of the Moss model, the lora t