v2.0 Curate

Hello everyone, this is the v2.0 Curate version of this LoRA.

各位好，这是这个 LoRA 的 v2.0 Curate 版本。

In this version, I reorganized the dataset and revised the captions compared with v1.

在这个版本中，我相比 v1 重新整理了数据集，并对标签进行了调整。

The amount of sketch-like or lineart-heavy influence has been reduced significantly.

相比上一版，这次已经减少了大量线稿风格的影响。

However, some images may still appear slightly overexposed in certain generations.

不过在部分出图中，仍然可能会出现轻微过曝的问题。

I strongly recommend adding “overexposed” to the negative prompt if you encounter this issue.

如果遇到这种情况，强烈建议在负面提示词中加入 “overexposed”。

This version is still not perfect, so any feedback or issue reports would be greatly appreciated.

这个版本依然不算完美，如果在使用中遇到任何问题，也欢迎反馈给我，我会非常感谢。

Trigger word: TNLmix2

触发词：TNLmix2

v1.0

Hello everyone, this is an experimental mixed-style LoRA mainly based on the Illustrious-XL model.

各位好，这是一个实验性的混合风格 LoRA，主要基于光辉模型制作。

Since this is my first attempt at using agent-assisted automatic captioning for training, the results may not be very stable.

由于这是我第一次尝试使用 agent 自动打标并进行训练，因此出图效果可能还不够稳定。

Some generations may show a strong blue or cyan color cast.

部分出图可能会出现较明显的蓝色或青色滤镜偏色。

You may try adding “blue tint, cyan tint, blue color cast, monochromatic blue” to the negative prompt to reduce this issue, although the result may not always be perfect.

可以尝试在负面提示词中加入 “blue tint, cyan tint, blue color cast, monochromatic blue” 来缓解这个问题，但效果可能不一定完全理想。

Overall, this LoRA should be considered an experimental release rather than a polished final version.

总体来说，这个 LoRA 更偏向一次实验性发布，并不是非常成熟的最终版本。

I plan to reorganize the dataset and captions myself in the future and release an improved version later.

后续我会重新整理数据集和标签，并尝试发布更加完善的升级版本。

If you encounter any issues or have suggestions, I would greatly appreciate your feedback.

如果在使用过程中遇到任何问题，或有任何建议，也欢迎反馈给我，我会非常感谢。

Trigger word: TNLmix1

触发词：TNLmix1

Recommended weight: 0.7–1.0

推荐权重：0.7–1.0

Recommended samplers: dpmpp_2m, er_sde

推荐采样器：dpmpp_2m, er_sde