AISHELL-3 original human recordings
Copyright Beijing Shell Shell Technology Co., Ltd.
Apache License 2.0; see LICENSE.txt
https://www.openslr.org/93/
https://huggingface.co/datasets/AISHELL/AISHELL-3
Revision: f20d5db4a31fe779ef07bb1af4ea92da5c786622
Corpus authors: Yao Shi, Hui Bu, Xin Xu, Shaoji Zhang, Ming Li.
Speaker labels are anonymous corpus identifiers, not BaoBaoGo voice actors.
BaoBaoGo modifications: constant-gain loudness normalization (target RMS -20 dBFS, peak <= -1 dBFS), MP3 encoding, sentence punctuation, Pinyin display, original Vietnamese translations. Timing and pitch preserved. No voice cloning or speech synthesis.
Detailed source paths and original WAV/derived MP3 hashes: ../data/human-examples.json
Encoding tool lameenc/LAME used locally, not distributed in app.