This repo contains a light-weight 8-bit encoded version of the original Voice Cloning ToolKit (VCTK) dataset, which is directly adapted from an official subset, Device Recorded VCTK (DR-VCTK) except that:
- 16-bit data are encoded into 8-bit through mu-law transformation
- silence parts (below tenth of maximum magnitude) are trimmed from the start and the end of each track
please check preprocess_demo.py
for details
Device Recorded VCTK (Small subset version)
RELEASE November 2017
National Institute of Informatics (NII)
JAPAN
Copyright (c) 2017
The Centre for Speech Technology Research (CSTR)
University of Edinburgh
UK
Copyright (c) 2017
Dr. Junichi Yamagishi