ACE-Step-v1.5-Kawaii_Future_Bass-LoRA

这是一个由580首 Kawaii Future Bass 风格音乐数据集训练的 LoRA 模型。该模型擅长生成欢快、充满活力的Kawaii Future Bass 音乐。

This LoRA model is trained on a Kawaii Future Bass dataset, specializing in generating upbeat, energetic Future Bass music.Trained on 580 Kawaii Future Bass music.

触发词 / Trigger Word

请在提示词中使用 Kawaii_Future_Bass 作为风格触发词。

Use Kawaii_Future_Bass in your prompt to trigger the style.

使用建议 / Usage Tips

语言与歌词 / Language & Lyrics

由于训练数据集主要由**日语歌曲**和**纯音乐**组成：

- 日语生成效果最佳：生成日语歌词时效果最自然。

- 其他语言注意事项：如果使用英语或其他语言的歌词，可能会出现咬字不清或发音奇怪的情况。

- 纯音乐生成：数据集中约 34% 为纯音乐，因此该模型也非常适合生成

曲目。

Since the training dataset consists mainly of Japanese songs and Instrumental tracks:

- Japanese lyrics work best: The model produces the most natural results with Japanese lyrics.

- Other languages: Generating lyrics in English or other languages may result in unclear pronunciation or artifacts.

- Instrumental tracks: About 34% of the dataset is instrumental, making this model excellent for generating instrumental tracks as well.

权重调整 / Weight Adjustment

- 默认权重1.0 (适用于日语或纯音乐生成)

- 非日语歌词：建议适当**降低 LoRA 权重**（例如 0.6 - 0.8）以改善咬字清晰度。

- Default Weight: 1.0 (Recommended for Japanese or Instrumental generation)

- Non-Japanese Lyrics: It is recommended to lower the LoRA weight (e.g., 0.6 - 0.8) to improve pronunciation clarity.

推荐参数 / Recommended Parameters

为了获得最佳效果，建议使用以下参数设置：

For best results, we recommend the following settings:

| 参数 (Parameter) | 推荐值 (Value) | 说明 (Note) |

| BPM | 87 | 数据集中最常见的速度 (Most common BPM in dataset) |

| Time Signature | 4/4 | 绝大多数样本为 4/4 拍 (Dominant time signature) |

| Key | D Major - E Major | 推荐 D 大调或 E 大调 (Recommended keys) |

| CFG Scale | 2.0 - 3.0 | 引导系数 (Guidance scale) |