中文简介

这是一个基于 Krea 2 训练的 LoRA 模型，训练集主要使用此前整理的 NAI3 素材，重点保留并强化 NAI3 风格下的人物表现、构图、色彩以及整体视觉质感。

模型支持自然语言提示词，但更推荐使用 Tag 形式进行生成。使用清晰、独立的 Tag 通常可以获得更稳定、更容易控制的结果，尤其适合人物、服装、姿势、镜头、光影和画风等要素的组合。

推荐提示词结构：

主体, 人物特征, 服装, 动作, 构图, 镜头, 光影, 风格

例如：

1girl, solo, long hair, black dress, looking at viewer, dynamic pose, upper body, cinematic lighting, rim light, detailed eyes, high detail

交流群q群451450868 discord Discord

English Description

This is a LoRA model trained for Krea 2, using a dataset primarily based on previously collected NAI3 materials. The model is designed to preserve and strengthen the character rendering, composition, color language, and overall visual aesthetics commonly associated with the NAI3 style.

The model supports natural-language prompts, but Tag-based prompting is highly recommended. Clear and separated tags generally provide more stable and controllable results, especially when combining character features, outfits, poses, camera angles, lighting, and stylistic elements.

Recommended prompt structure:

subject, character features, outfit, pose, composition, camera, lighting, style

Example:

1girl, solo, long hair, black dress, looking at viewer, dynamic pose, upper body, cinematic lighting, rim light, detailed eyes, high detail

交流群q群451450868 discord Discord