metadata
license: wtfpl
language:
- en
size_categories:
- n<1K
dataset_info:
features:
- name: wav_file
dtype: audio
- name: text
dtype: string
splits:
- name: train
num_bytes: 183699060
num_examples: 648
download_size: 183379554
dataset_size: 183699060
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
These are voice clips from Yanfei from Genshin Impact, taken from the simon3000 genshin voice dataset, re-transcribed with whisper, to remove the weird transcription and vocal effects, sampled to 22.05khz sampling rate, run through simple noise reduction with librosa, and the longer entries were chunked to be up to 15 seconds long, and the silent parts were trimmed, also the very small unintelligible audios that couldn't even be transcribed were scrapped from the set. Finally my plan is to run it through the dataspeech recipe to make it ready to be used with parlerTTS for finetuning a Yanfei voice.