zldzmfoq12 / VCtube

A pakage for crawling audio from Youtube

Audio, Caption Crawler and Processor -TTS Data Generator-

Downloads and processes the audios and captions(subtitles) from Youtube videos for Speech AI

Generates audio datas from Youtube for TTS

You can get more details in http://www.interspeech2020.org/index.php?m=content&c=index&a=show&catid=392&id=516

Requirements

• Currently requires python >= 3.6
• FFmpeg
• youtube_dl
• pydub
• youtube_transcript_api

To Use

  pip install vctube

---

  from vctube import VCtube

  playlist_name=""
  playlist_url = ""
  lang = ""   #ex) ko, en, fr, de...

  vc = VCtube(playlist_name, playlist_url, lang)

  vc.download_audio()    #download audios from youtube

  vc.download_captions()  #download captions from youtube

  vc.audio_split()       #split audio with captions

Results

  datasets
    |- playlist name
        |- metadata.csv
        |- alignment.json
        |- wavs
             ├── 1.wav
             ├── 2.wav
             ├── 3.wav
             └── ...

and metadata.csv should look like:

{
    "0001.wav|그래서 사람들도 날 핍이라고 불렀다.",
    "0002.wav|크리스마스 덕분에 부엌에 먹을게 가득했다.",
    "0003.wav|조가 자신이 그 사람이라고 나섰다.",
    ...
}

and alignment.json should look like:

{
    "./datasets/playlist name/wavs/0001.wav": "그래서 사람들도 날 핍이라고 불렀다.",
    "./datasets/playlist name/wavs/0002.wav": "크리스마스 덕분에 부엌에 먹을게 가득했다.",
    "./datasets/playlist name/wavs/0003.wav": "조가 자신이 그 사람이라고 나섰다.",
    ...
}

Pypi address

https://pypi.org/project/vctube/