alna12fal-style-loha

Artist Style LoRA

• Artist: あるな
• Pixiv:https://www.pixiv.net/users/16594489/artworks

Samples

中文

这是一个基于 Anima 训练的画风模型项目。

本仓库保留了同一训练过程中的多个检查点。

不同轮数可能在画风强度、泛化能力和细节表现之间有所差异。

欢迎根据个人需求自行测试和选择。

项目主要关注：

• 画师风格学习
• 光影表现
• 皮肤质感
• 整体审美还原

训练过程中更注重学习画风本身，而非角色身份。

触发词

@jk348

推荐提示词

newest, masterpiece, best quality

推荐权重

0.7 ~ 1.0

训练信息

• Base Model：Anima
• Optimizer：AdamW
• Scheduler：Cosine
• 删除角色名称标签
• 删除部分固定外观标签
• 保留场景、服装、动作等内容标签
• 目标为学习画风而非角色身份

---

免责声明

本仓库仅用于研究、学习和画风训练实验。

原始作品版权归对应创作者所有。

如您为相关作品权利人并对本仓库内容有任何疑问，请联系我，我会及时处理。

---

English

Introduction

This repository contains an artist style model trained on Anima.

This repository contains multiple checkpoints from the same training run.

Different checkpoints may offer different balances between style strength and flexibility.

Feel free to test and choose the version that best fits your workflow.

The project mainly focuses on:

• Artist style learning
• Lighting and shading
• Skin rendering
• Overall aesthetic reproduction

The training process focuses on learning artistic style rather than character identity.

Trigger Word

@jk348

Recommended Prompt

newest, masterpiece, best quality

Recommended Weight

0.7 ~ 1.0

Training Information

• Base Model: Anima
• Optimizer: AdamW
• Scheduler: Cosine
• Character tags removed
• Partial appearance tags removed
• Scene, outfit and pose tags retained
• Focused on style learning rather than character identity

---

Disclaimer

This repository is intended for research, learning, and style transfer experiments.

All rights to the original artworks belong to their respective creators.

If you are the copyright holder of the referenced artworks and have concerns regarding this repository, please contact me and I will address the issue appropriately.