Skip to content

Commit

Permalink
feat: add audio component support
Browse files Browse the repository at this point in the history
  • Loading branch information
rockbenben committed Oct 11, 2023
1 parent b39cf13 commit 1407dfd
Show file tree
Hide file tree
Showing 2 changed files with 2 additions and 4 deletions.
2 changes: 1 addition & 1 deletion docs/.vuepress/theme.ts
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,7 @@ export default hopeTheme({

// 组件库
components: {
components: ["Badge", "BiliBili", "VideoPlayer", "YouTube"],
components: ["Badge", "BiliBili", "AudioPlayer", "VideoPlayer", "YouTube"],
},

// 禁用不需要的配置
Expand Down
4 changes: 1 addition & 3 deletions docs/_posts/2023-10-07-clone-voice.md
Original file line number Diff line number Diff line change
Expand Up @@ -83,9 +83,7 @@ order: -56

### 语音克隆的「口音」问题

> 每个人的声音都是独一无二的,克隆自己的声音可以用于制作高度个性化的内容。
示例音频:<AudioPlayer src="https://oss.newzone.top/audio/clonevoice01.wav" title="克隆音频实例 2" />
<AudioPlayer src="https://oss.newzone.top/audio/clonevoice01.wav" title="克隆音频实例 2" />

这个示例是使用 8 分钟 B 站视频和 CJE 模型训练出的。但你可能注意到了明显的断调口音问题,仿佛一个日本人在说中文。正如 @zachx121 指出的,「CJE 用的时候 romaji 的注音,就好比说用汉语拼音去标注英文单词的发音一样会有“口音”」。为了避免这个问题,可以使用纯中文设计的 C 模式进行训练和生成,以确保音频的自然和准确性。增加训练次数也有助于改善口音问题。

Expand Down

0 comments on commit 1407dfd

Please sign in to comment.