ifmanzhang/huangquan-audio-training-data

Huangquan Audio Training/Reference Data

This archive contains external speech training/reference material used in the Huangquan Road character TTS workflow.

Contents

• 5002 FLAC files
• Six speaker directories: p227, p238, p245, p281, p318, p340
• Original archive: Audio_Training_Data.zip
• Compressed archive size: approximately 634.3 MB
• SHA256: CAC56B2E327FA6E5D1E06065CE94B1A68BCC5259B8BF930842F0BDE8351BD0BF

Source And Licence

The audio source is the CSTR VCTK Corpus version 0.92:

Yamagishi, Junichi; Veaux, Christophe; MacDonald, Kirsten. (2019). CSTR VCTK Corpus: English Multi-speaker Corpus for CSTR Voice Cloning Toolkit (version 0.92), [sound]. University of Edinburgh. The Centre for Speech Technology Research (CSTR). https://doi.org/10.7488/ds/2645.

The VCTK README states that the corpus is licensed under Creative Commons Attribution 4.0 International.

Project Use

This archive is used as external VCTK-derived training/reference audio for the project's TTS workflow. It should not be described as project-original recorded speech.

Related Project Resources

• Main project repository: https://git.arts.ac.uk/y-zhang0320251/MML_Final.git
• Project datasheet: see DATASHEET.md in the main repository.
• Technical notes: see docs/TECHNICAL_HIGHLIGHTS.md in the main repository.