jinluan lora | 金鸾 lora 细节增强/画质改善 | Detail enhancement/image quality improvementDetail enhancement/image quality improvement

详情

下载文件

模型描述

转载模型前必须标明原模型地址,并征得我同意后方可转载,否则我将向目标平台投诉并要求下架,感谢合作

承接LoRA模型训练,QQ:3032781264/留言或私聊 | Refining Lora model, QQ: 3032781264/Leave a message or private chat

该模型经多人测试,具备良好的质量调整与细节优化能力

只需添加触发词后,再附加起手式或其他质量标签,即可获得优异的画质调整效果

建议权重调整至0.8以下,并配合

JRU(The journey remains unfinished)道途未尽2,效果更为显著

After being tested by multiple people, this model has the ability to adjust quality and improve details effectively

Simply add a trigger word and then add a hand gesture or other quality tag to achieve good image quality adjustment

Adjust the weight to below 0.8 in coordination

JRU (The Journey Remains Unfinished) has a more noticeable effect when the journey is incomplete

金鸾 lora

--------------------------------------------------------------------------------

训练图片:12张

训练基础模型:JRU(The journey remains unfinished)道途未尽2
训练步数:10批次,3500步

--------------------------------------------------------------------------------

建议:

Checkpoint:以下为我个人推荐:

JRU(The journey remains unfinished)道途未尽2
JRU(The journey remains unfinished)道途未尽2 - v2.0XL |Stable Diffusion XL 检查点 |奇维泰 (civitai.com)

Deep Dark Hentai Mix - NSFW Anime

权重:建议使用0.6至1.0之间,经多次测试,0.7为最合适值,具体视个人希望生成的图像类型而定。

采样器:DPM++SDE Karras,20至30步。

采样器:DPM++2M Karras,20至30步。
推荐标签:(yellow theme,yellow dress,black hair,forehead mark)

LoRA触发词:jinluan

训练集作者出处
https://www.pixiv.net/artworks/119394380

此模型生成的图像

未找到图像。