Anima 版本建议使用精简的质量提示词
It is recommended to use streamlined quality prompts for the Anima version.
免责声明
本 LoRA 为《王者荣耀》相关角色/皮肤的同人模型,仅供个人学习、研究、交流与非商业同人创作使用。相关角色与美术形象的版权归原版权方所有。本人不拥有相关版权,本 LoRA 亦非官方内容,未获得官方授权。禁止将本 LoRA 或使用本 LoRA 生成的内容用于任何商业用途
Krea2:
推荐 LoRA 强度:1
Recommended LoRA weight: 1
推荐步数:8-10
Recommended sampling steps: 8–10
推荐 CFG:1
Recommended CFG scale: 1
推荐采样器: er_sde,Euler
Recommended samplers: er_sde, Euler
Anima:
推荐 LoRA 强度:0.9
Recommended LoRA strength: 0.9
推荐步数:30
Recommended steps: 25
推荐 CFG:5
Recommended CFG: 5
推荐采样器:Euler A
Recommended samplers: Euler A
这是一个角色 LoRA,角色为《王者荣耀》中朵莉亚的皮肤「幻珠鲛人」。
该版本仅训练了朵莉亚的人形态,不包含人鱼形态。
如果您发现该 LoRA 与您使用的画风提示词或其他 LoRA 兼容性不佳,欢迎向我反馈。我会尝试制作泛化版,但不保证一定能够完全成功。
触发词:Dolia_HZJR
建议搭配以下角色特征、服装与主题提示词使用:1girl, solo, white hair, very long hair, double bun, pink hair, gradient hair, multicolored hair, ahoge, hair ornament, flower hair ornament, golden hair ornament, jewelry, earrings, bare shoulders, pink clothes, white clothes, purple clothes, gold trim, ornate clothes, chinese clothes, fantasy clothes, ruqun, layered dress, long skirt, translucent sleeves, translucent fabric, shawl, long ribbons, flowing ribbons, flower-shaped pearl, pearl, petals, floating petals, flowers, butterflies, glowing particles, water, fish, jellyfish, lotus, wisteria
由于训练数据和模型表现限制,部分服装结构、双髻发型、花瓣头饰、金饰、透明鲛绡、飘带及鲛珠细节可能无法始终完整还原,还请见谅。
请注意,本版本并未训练人鱼形态,因此不建议加入 mermaid、fish tail 等提示词。
如果在使用过程中遇到任何问题,欢迎随时向我反馈。
非常感谢大家的支持!
This is a character LoRA for Dolia’s “Huanzhu Jiaoren” skin from Honor of Kings.
This version was trained exclusively on Dolia’s human form and does not include her mermaid form.
If you find that this LoRA does not work well with your preferred style prompts or other LoRAs, please feel free to let me know. I may try to create a more generalized version, although I cannot guarantee that it will be completely successful.
Trigger word: Dolia_HZJR
Recommended character, outfit, and thematic tags: 1girl, solo, white hair, very long hair, double bun, pink hair, gradient hair, multicolored hair, ahoge, hair ornament, flower hair ornament, golden hair ornament, jewelry, earrings, bare shoulders, pink clothes, white clothes, purple clothes, gold trim, ornate clothes, chinese clothes, fantasy clothes, ruqun, layered dress, long skirt, translucent sleeves, translucent fabric, shawl, long ribbons, flowing ribbons, flower-shaped pearl, pearl, petals, floating petals, flowers, butterflies, glowing particles, water, fish, jellyfish, lotus, wisteria
Due to limitations in the training data and model performance, some clothing structures, double-bun hairstyle details, petal-shaped hair ornaments, golden accessories, translucent merfolk silk, ribbons, and pearl details may not always be reproduced perfectly. Thank you for your understanding.
Please note that this version was not trained on her mermaid form, so tags such as mermaid and fish tail are not recommended.
If you encounter any problems while using this LoRA, please feel free to contact me at any time.
Thank you very much for all your support!
Description
关于 anima 版本,我不知道我的逻辑是否正确:LoRA 的训练文本越密集、越接近自然语言、包含的视觉描述越完整,推理时就越容易与冗长的画风 / 质量描述发生语义叠加;此时简洁而明确的质量词通常更稳定。 这就意味着,如果您使用更加简洁精确的质量提示词,效果反而会更好
Regarding the Anima version, I’m not sure if my reasoning holds true: the denser the training captions for a LoRA are, the closer they are to natural language, and the more comprehensive the visual descriptions they contain, the more prone they are to semantic overlap with lengthy stylistic and quality descriptors during inference. In such cases, concise, explicit quality keywords tend to yield far more consistent results.
This implies that using shorter, precise quality prompts will actually deliver superior outcomes.
FAQ
Comments (1)
还是躲在水里最自在









