metadata

dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: rttm
      dtype: string
  splits:
    - name: train
      num_bytes: 3326774904
      num_examples: 126
  download_size: 3323845215
  dataset_size: 3326774904
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
language:
  - pt
task_categories:
  - audio-to-audio
tags:
  - fudencio
  - RVC
  - voice
  - Brazil
  - Brasil
  - Portugues
  - português
  - south
  - park
size_categories:
  - n<1K

What is this?

This is a dataset extracted from brazilian cartoon called Fudêncio (by MTV), which is somehow similar to South Park.

This dataset has three features

rttm string ( to identify speakers ), episode_name ( for reference ) and the audio voicelines Acapella made using Demucs (no_voice files are not included)

The separation of Demucs is not perfect, but can help people to train RVC voices for different characters.

Plan to future upgrades

I plan to use this later to split voicelines by characters names into another dataset.