https://github.com/adkcodexd/anime-audio-dataset-maker
Anime audio speaker recognize and classify. A easy python script for vits data set make.
https://github.com/adkcodexd/anime-audio-dataset-maker
Last synced: 12 months ago
JSON representation
Anime audio speaker recognize and classify. A easy python script for vits data set make.
- Host: GitHub
- URL: https://github.com/adkcodexd/anime-audio-dataset-maker
- Owner: ADKcodeXD
- License: mit
- Created: 2023-12-14T16:40:14.000Z (over 2 years ago)
- Default Branch: master
- Last Pushed: 2024-01-27T15:00:52.000Z (over 2 years ago)
- Last Synced: 2025-01-26T16:44:16.686Z (over 1 year ago)
- Language: Python
- Homepage:
- Size: 28.6 MB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Anime Audio DataSet Maker
- [English README](README_en.md)
- [中文说明](README.md)
## Introduce
此项目旨在为番剧提供一个快速高效提取角色音频的解决方案。
WEBUI下载链接：
Anime-Audio-Dataset-Maker-WEBUI Release
## 安装&使用
- 第一个方法 整合包一键下载使用方法:
链接: https://pan.baidu.com/s/1T9GbDo6enrV__G0j7pXbwQ?pwd=s556 提取码: s556
下载后使用 整合包使用这个.bat 即可
- 安装使用 首先先安装pytorch，
这个需要根据系统的cuda版本来进行安装
以我的Cuda11.8为例
```sh
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
```
在这里pytorch 选择你对应的版本并运行相对应的命令即可~
- 然后安装该仓库需要的依赖
```sh
pip3 install -r requirement.txt
```
- 下载webui
Anime-Audio-Dataset-Maker-WEBUI Release
下载最新版本的webui解压至该项目根目录下
- 运行launch.bat
项目会运行在7896端口
## How it work
- 通过pyannote.audio对原音频进行说话人的识别和切割
- 通过字幕时间线对原音频进行切割
- 通过匹配检测最佳匹配的说话人
- 分类到各个说话人的文件夹中
## WebUI操作流程
- 开始预处理音频
...
## Feature
- Support automaticly split long audio by each speaker
- Support sub upload and slice by sub timeline.
- Support edit the sub text and export it by bert-vits config
- Support split ever single audio (WebUI)
- Support merge audio with interval (WebUI)
- Support management folders or files (WebUI)
- Support use Arrow key to handle data (WebUI)
- Support batch rename (WebUI)
- Support batch move or remove (WebUI)