Eldreda Sunbringer｜Anima Character LoRA

Eldreda Sunbringer from Zenless Zone Zero — her tall, adult form.

A character LoRA trained from Anima Base, focusing on her silver-white hair, distinctive headpiece, asymmetrical outfit, and black cape with blue lining. The dataset includes both caped and capeless appearances.

This version uses individually reviewed eye-color captions: her right eye is golden yellow, and her left eye is cyan blue.

Trigger Word

gb_eldreda

Adjust the cape using prompts: use black cape for the caped look, or try without cape, bare shoulders for the capeless look.

For more consistent results, include her hair color, eye colors, headpiece, and outfit details alongside the trigger.

Suggested Settings

• Base model: Anima Base v1.0

• Starting strength: 0.8; compare 0.6–1.0

• Resolution: 832×1216 or a similar pixel count

• Sampling steps: 28

• CFG: 4.5

The strength range is a starting point for testing, rather than a verified optimum. Use your usual Anima-compatible workflow.

Training Details

• Trainer: AI Toolkit

• Dataset: 40 curated images, including 13 images with AI-assisted background cleanup

• Training steps: 1,500

• Rank / Alpha: 32 / 32

• Training resolution: 1024 bucketing

• Horizontal flipping: Disabled to preserve eye-color orientation

Notes

In the fixed-seed training previews, the detailed portrait prompt produced the intended eye colors. The trigger alone does not consistently reproduce her hair color, heterochromia, or headpiece. Headpiece proportions and fine clothing details may vary.

艾尔妲·挽昼｜Anima 角色 LoRA

来自《绝区零》的 艾尔妲·挽昼，以大挽昼形态为训练目标。

基于 Anima Base 训练，主要学习银白长发、标志性头饰、不对称服装，以及蓝色内衬的黑色披风。训练素材同时包含带披风与无披风造型。

本版本逐图复核了眼睛颜色标注：角色自身右眼金黄、左眼蓝青，即正面画面左金右蓝。

触发词

gb_eldreda

可通过提示词调整披风状态：使用 black cape 生成带披风造型，使用 without cape, bare shoulders 尝试无披风造型。

建议同时补充白发、左右眼、头饰及服装描述，以获得更稳定的效果。

使用建议

• 底模： Anima Base v1.0

• 起步权重： 0.8，可对比 0.6～1.0

• 分辨率： 832×1216 或相近像素量

• 采样步数： 28

• CFG： 4.5

权重范围仅作为试用起点，尚未验证最佳值。使用你现有的 Anima 兼容工作流即可。

训练信息

• 训练器： AI Toolkit

• 训练素材： 40 张筛选图片，其中 13 张经过生成式背景清理

• 训练步数： 1500

• Rank / Alpha： 32 / 32

• 训练分辨率： 1024 分桶

• 水平翻转： 关闭，避免左右眼颜色颠倒

已知限制

在固定种子的训练预览中，明确描述左右眼的肖像提示能够生成正确眼色。只使用触发词时，发色、异色瞳和头饰仍不够稳定；头饰比例及服装细节也可能出现偏差。

示例提示词 / Example Prompts

带披风肖像 / Portrait with cape

gb_eldreda, 1girl, solo, long white hair, heterochromia,
her right eye is golden yellow and her left eye is cyan blue,
black horn-shaped headpiece, front view, portrait,
both eyes open, looking at viewer,
black cape with blue lining, red collar,
soft daylight, garden background

无披风坐姿 / Seated, without cape

gb_eldreda, 1girl, solo, long white hair, heterochromia,
her right eye is golden yellow and her left eye is cyan blue,
black horn-shaped headpiece, dark blue and black bodysuit,
bare shoulders, black gloves, asymmetrical legwear,
without cape, sitting on a park bench, daylight

服饰提示词/Sample outfit prompts

gb_eldreda, 1girl, solo, long silver-white hair, curled sidelocks,
heterochromia, her right eye is golden yellow and her left eye is cyan blue,
black mechanical headpiece, long black earrings,
black cape with red-lined standing collar,
asymmetrical navy blue and black bodysuit, curved gold trim

负面提示词 / Negative Prompt

worst quality, low quality, blurry, jpeg artifacts,
text, watermark, logo, signature, speech bubble, comic panel

角色及原始素材归原权利方所有。
The character and original source materials belong to their respective rights holders.