litagin's picture
Update README.md
386974f verified
metadata
license: other
task_categories:
  - automatic-speech-recognition
  - text-to-speech
language:
  - ja
pretty_name: ReazonSpeechV2Denoised
size_categories:
  - 10M<n<100M

Reazon Speech v2 DENOISED

  • Reazon Speech v2の音声ファイルをUVRを使用してBGMやノイズ除去したこのデータセットのミラーです。
  • 作者(のおそらく一人?)であるStardust-minusさんによるデータセットの説明:

    Reazon-Speech2データセットは、現在UVRによって処理され公開されています。私は3674/4096ファイルの処理を完了しました。全データセットは8台のA800 GPUを使用して約10日間でクリーニングされました。

English ver

  • This is a mirror of this dataset, which consits of audio files of Reazon Speech v2 denoised and background music removed using UVR.
  • Dataset description by (maybe? one of) the author(s) Stardust-minus:

    The Reazon-Speech2 Dataset, which have been ran UVR now public. I have completed about 3674/4096 files. The whole dataset was cleaned on 8*A800 GPUs about ~10days.

Original Dataset Source

  • All the credits go to the original authors and contributors.

Reazon Speech v2

Denoised ver

License

This dataset is licensed under the same license as the original Reazon Speech corpus: CDLA-Sharing-1.0 with the following restriction:

TO USE THIS DATASET, YOU MUST AGREE THAT YOU WILL USE THE DATASET SOLELY FOR THE PURPOSE OF JAPANESE COPYRIGHT ACT ARTICLE 30-4.

Disclaimer (from the original Reazon Speech)

TO USE THIS DATASET, YOU MUST AGREE THAT YOU WILL USE THE DATASET SOLELY FOR THE PURPOSE OF JAPANESE COPYRIGHT ACT ARTICLE 30-4.

Dataset Format

  • No transcripstions, so please refer to e.g. all.tsv of this mirror for the transcriptions of audio files.
  • No modifications of audio file names.

Structure of the repository (note that it's different from the Reazon Speech, and be aware of the directory data/001.tar/)

.
└── data
    ├── 001.tar/
    │   └── 001.tar
    ├── 002.tar/
    │   └── 002.tar
    ...