Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -14,22 +14,62 @@ pretty_name: genshin_voice_sovits
|
|
14 |
|
15 |
本仓库用于预览训练出的各种语音模型的效果,点击角色名自动跳转对应训练参数。</br>
|
16 |
正常说话的音色转换较为准确,歌曲包含较广的音域且bgm和声等难以去除干净,效果有所折扣。</br>
|
17 |
-
|
18 |
-
<pre>
|
19 |
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
<
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
关键参数:
|
30 |
|
31 |
-
|
32 |
-
epoch:
|
33 |
|
34 |
其余:
|
35 |
batch_size = 一个step训练的片段数<br>
|
@@ -42,5 +82,5 @@ step=segments*epoch/batch_size,即模型文件后面数字由来<br>
|
|
42 |
差不多2800epoch(10000step)就已经出结果了,实际使用的是5571epoch(19500step)的文件,被训练音色和原音色相差几<br>
|
43 |
何,差不多有个概念。当然即使loss也不足以参考,唯一的衡量标准就是当事人的耳朵。当然,正常训练,10min还是有些少的。<br>
|
44 |
|
45 |
-
|
46 |
![sanbing_loss](./散兵效果预览/%E8%AE%AD%E7%BB%83%E5%8F%82%E6%95%B0%E9%80%9F%E8%A7%88.assets/sanbing_loss.png)
|
|
|
14 |
|
15 |
本仓库用于预览训练出的各种语音模型的效果,点击角色名自动跳转对应训练参数。</br>
|
16 |
正常说话的音色转换较为准确,歌曲包含较广的音域且bgm和声等难以去除干净,效果有所折扣。</br>
|
17 |
+
|
|
|
18 |
|
19 |
+
<div style="width: 800px; height: 300px; overflow: auto;">
|
20 |
+
<table border="1" style="white-space: nowrap; text-align: center;">
|
21 |
+
<thead>
|
22 |
+
<tr>
|
23 |
+
<th>角色名</th>
|
24 |
+
<th>角色原声A</th>
|
25 |
+
<th>被转换人声B</th>
|
26 |
+
<th>A音色替换B</th>
|
27 |
+
<th>A音色翻唱(点击直接下载)</th>
|
28 |
+
</tr>
|
29 |
+
</thead>
|
30 |
+
<tbody>
|
31 |
+
<tr>
|
32 |
+
<td><a href="https://huggingface.co/datasets/jiaheillu/audio_preview/blob/main/散兵效果预览/训练参数速览.md">散兵</a></td>
|
33 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/散兵效果预览/部分训练集/真遗憾,小吉祥草王让他消除了那么多的切片,剥夺了我将他一片一片千刀万剐的快乐%E3%80%82.mp3" controls="controls"></audio></td>
|
34 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/散兵效果预览/原声/shenli3.wav" controls="controls"></audio></td>
|
35 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/散兵效果预览/转换结果/shenli3mp3_auto_liulangzhe.wav" controls="controls"></audio></td>
|
36 |
+
<td><a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/散兵效果预览/转换结果/夢で逢えたら2liulangzhe_f.wav">夢で会えたら</a></td>
|
37 |
+
</tr>
|
38 |
+
<tr>
|
39 |
+
<td><a href="https://huggingface.co/datasets/jiaheillu/audio_preview/blob/main/胡桃_preview/README.md">胡桃</a></td>
|
40 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/%E8%83%A1%E6%A1%83_preview/hutao.wav" controls="controls"></audio></td>
|
41 |
+
<td>.........</td>
|
42 |
+
<td>.........</td>
|
43 |
+
<td>
|
44 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/胡桃_preview/moonlight_shadow2胡桃.WAV">moonlight shadow</a>,
|
45 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/胡桃_preview/云烟成雨2胡桃.WAV">云烟成雨</a>,
|
46 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/胡桃_preview/原点2胡桃.WAV">原点</a>,
|
47 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/胡桃_preview/夢だ会えたら2胡桃.WAV">夢で逢えたら</a>,
|
48 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/胡桃_preview/贝加尔湖畔2胡桃.WAV">贝加尔湖畔</a>
|
49 |
+
</td>
|
50 |
+
</tr>
|
51 |
+
<tr>
|
52 |
+
<td><a href="https://huggingface.co/datasets/jiaheillu/audio_preview/blob/main/绫华_preview/README.md">神里绫华</a></td>
|
53 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/linghua428.wav" controls="controls"></audio></td>
|
54 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/yelan.wav" controls="controls"></audio></td>
|
55 |
+
<td><audio src="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/yelan.wav_auto_linghua_0.5.flac" controls="controls"></audio></td>
|
56 |
+
<td>
|
57 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/アムリタ2绫华.WAV">アムリタ</a>,
|
58 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/大鱼2绫华.WAV">大鱼</a>,
|
59 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/遊園施設2绫华.WAV">遊園施設</a>,
|
60 |
+
<a href="https://huggingface.co/datasets/jiaheillu/audio_preview/resolve/main/绫华_preview/the_day_you_want_away2绫华.WAV">the day you want away</a>
|
61 |
+
</td>
|
62 |
+
</tr>
|
63 |
+
</tbody>
|
64 |
+
</table>
|
65 |
+
</div>
|
66 |
+
|
67 |
+
|
68 |
|
69 |
关键参数:
|
70 |
|
71 |
+
audio duration:训练集总时长
|
72 |
+
epoch: 轮数
|
73 |
|
74 |
其余:
|
75 |
batch_size = 一个step训练的片段数<br>
|
|
|
82 |
差不多2800epoch(10000step)就已经出结果了,实际使用的是5571epoch(19500step)的文件,被训练音色和原音色相差几<br>
|
83 |
何,差不多有个概念。当然即使loss也不足以参考,唯一的衡量标准就是当事人的耳朵。当然,正常训练,10min还是有些少的。<br>
|
84 |
|
85 |
+
[点我查看相关文件](https://huggingface.co/datasets/jiaheillu/audio_preview/tree/main)<br>
|
86 |
![sanbing_loss](./散兵效果预览/%E8%AE%AD%E7%BB%83%E5%8F%82%E6%95%B0%E9%80%9F%E8%A7%88.assets/sanbing_loss.png)
|