Model info

v1.0 has the highest fidelity.

This version is fine-tuned for WAI-IllustriousXL. However, the one trained on standard Illustrious XL is the most stable choice to avoid feature conflicts.↓↓↓

https://civitai.com/models/2314075/evanescia-or-detailed-honkai-star-rail

This LoRA is fine-tuned on a dataset of 550+ images. It can recreate character using the trigger word Evanescia alone, while also providing detachable clothing tags.

In addition to LoRA based on WAI-IllustriousXL, I have also trained versions based on NoobAI, IllustriousXL, Anima, AnimagineXL, PonyXL.

Param Settings

• Recommended weight: 0.8 - 1.0

• example prompt :(Please adjust the prompts according to model card of the checkpoint you are using.)

masterpiece, best quality, very aesthetic, newest, hires, Evanescia, 1girl, solo, looking at viewer, smile, white jacket, white shirt, bow, upper body, detailed light, simple background,

• full body:

masterpiece, best quality, very aesthetic, newest, hires, Evanescia, 1girl, solo, smile, looking at viewer, white jacket, white shirt, bow, pantyhose, boots, detailed light, simple background,

For full-body shots, you can try adding full body or boots.

Tips

• Official design is bound to the following tags: white jacket, white shirt, skirt, bow, pantyhose, boots, gloves

• If you find the official outfit elements are not decoupling when using your own checkpoint, add cloth tags to the negative prompt. For example, if you want an bare-shoulder look, add bare shoulders to the positive prompt, remove white jacket or add it to the negative prompt.

Others

(Models are automatically deployed via AI Agent. Please report any parameter errors or version issues in the comments.)

While I am very satisfied with the accuracy of her features, please feel free to leave your generations or feedback.

Thank you for your support!

模型简介

v1.0有最高还原度

这版是在WAI-IllustriousXL 做的，但基于标准IllustriousXL训练的版本肯定是最稳妥的选择，因为不会有特征冲突。↓↓↓

https://civitai.com/models/2314075/evanescia-or-detailed-honkai-star-rail

550+ 张图片的数据集上微调得到。此LoRA使用触发词 Evanescia 即可还原人物，也提供了可拆卸的服装提示词，方便各类二创发挥。

除了基于WAI-IllustriousXL的版本，我还做了基于IllustriousXL, NoobAI, Anima, AnimagineXL, PonyXL版本。

推荐设置

• 推荐权重： 0.8 - 1.0

• （请根据你实际使用的Checkpoint模型的官方文档，适当调整提示词策略。）

• 上本身prompt:

masterpiece, best quality, very aesthetic, newest, hires, Evanescia, 1girl, solo, looking at viewer, smile, white jacket, white shirt, bow, upper body, detailed light, simple background,

• 全身prompt:

masterpiece, best quality, very aesthetic, newest, hires, Evanescia, 1girl, solo, smile, looking at viewer, white jacket, white shirt, bow, pantyhose, boots, detailed light, simple background,

全身可尝试添加full body或boots 。

Tips

• 官方服设绑定在以下标签：white jacket, white shirt, skirt, bow, pantyhose, boots, gloves

• 如果你用你的checkpoint推理的时候服装元素不能正确解耦，把上述服装tag加到负面提示词。举例子，如果想要光肩膀的效果，把bare shoulders 加入positive prompt，移除 white jacket或把它加进negative prompt。

小记

模型由AI Agent全自动发布，若参数、模型版本有问题，欢迎评论反馈。

虽然我已经对她的特征还原程度很满意，如果有任何反馈意见欢迎评论，以便我做出调整，谢谢。我会定期在社交媒体发布图片，可在对应平台取原图。

Bilibili账号：https://space.bilibili.com/12595237?spm_id_from=333.1007.0.0