This is a style LoRA trained on the art style of the game Dragon Traveller. The training dataset is composed of 231 high-quality official images from the game, including character stand-alone paintings and favorability plot illustrations, all of which are the original visual materials of the game to ensure the authenticity of the style reproduction.
这是一款基于游戏《龙魂旅人》美术风格训练的风格类LoRA，训练数据集由231张游戏官方高质量素材构成，包含角色立绘、好感度剧情插画等，均为游戏原版视觉素材，保证风格还原的真实性。
The LoRA performs excellently in light and shadow rendering and skin texture expression, and can restore the anime art style of Dragon Traveller well. The rendered skin has a translucent and delicate texture, and the light and shadow layering is natural and harmonious, which can add a unique game-style aesthetic to the generated images.
该LoRA在光影渲染和皮肤质感表现上表现出色，能较好还原《龙魂旅人》二次元美术风格，渲染出的皮肤具备通透细腻的质感，光影层次自然和谐，可为生成画面增添独特的游戏画风美感。
In the actual generation effect, there are certain differences between the character shape generated by the LoRA and the original characters in the game. This LoRA is mainly oriented to the reproduction of the game's art style rather than the precise restoration of the game's characters, which is a normal phenomenon of style-based LoRA training with game material.
在实际生成效果中，该LoRA生成的人物造型与游戏原版角色存在一定出入，本LoRA核心面向游戏美术风格的复刻，而非游戏角色的精准还原。
It is recommended to use the weight of 0.8-1.0 when using, (actually, you can decided your own weight freely and match your own taste),which can balance the style expression and the diversity of generated images. This LoRA is compatible with most mainstream 2D anime base models (such as prefectIllustriousXL_v5), and the generation effect is more stable under the model with semi-realistic style tendency.
使用时推荐搭配0.8-1.0的权重，(其实，你可以按自己的需求随意调整)可平衡风格表现与生成画面的多样性；该LoRA适配多数主流二次元基础模型（如prefectIllustriousXL_v5），在偏半写实风格倾向的模型下生成效果更稳定。
Suitable for creating 2D character portraits, game-style cover art, anime illustrations and other works. It can be used with scene, clothing and pose LoRAs to enrich the generated content, and the style fusion effect is good.
适用于创作二次元人物肖像、游戏风格封面、动漫插画等作品，可搭配场景、服饰、姿势类LoRA丰富生成内容，风格融合效果佳。
The training effect of the LoRA is acceptable in actual testing, and the follow-up will be optimized and updated according to the actual use feedback, including adjusting the training parameters and supplementing the training data to further improve the style expression and generation stability.
该LoRA经实际测试训练效果尚可，后续将根据实际使用反馈进行优化更新，包括调整训练参数、补充训练素材，进一步提升风格表现与生成稳定性。