Skip to main content
MENUCLOSE
Access &
Contact
日本語

Speech-rate conversion corpus: SpeedSpeech-JA-2022

What's new

  • Feb. 28, 2022 ver1.0.0 has been released.

Overview

This corpus is created in collaboration with System 1 lab., Graduate School of Information Science and Technology, the University of Tokyo. The main purpose of this corpus is to study the evaluation of speech-rate conversion in speech synthesis and voice conversion. This corpus consists of Japanese 324 sentences from the ITA corpus (RECITATION) read by one male and one female professional Japanese narrator at (1) normal speed, (2) fast speed, and (3) slow speed conditions, respectively. All the speech waveforms (24-bit linear PCM, sampling frequency: 48 kHz) were recorded in a soundproof room.

Download

speedspeech_ja_2022_v1.0.0
(about 1.36 GB)

License

This corpus is released under "Attribution 4.0 International (CC BY 4.0)".
The license of the text follows the ITA corpus license (public domain).

Extracted directory

The files have been compressed in 'zip' format. The extracted directory should look like:

-------------------------------------------------------------------------------------------
female/
  01_normal/  ##female normal spped 324 sentences (averaged speech-rate 4.8 mora / sec)
  02_fast/   ##female fast spped 324 sentences (averaged speech-rate 6.8 mora / sec)
  03_slow/   ##female slow spped 324 sentences (averaged speech-rate 3.8 mora / sec)
male/
  01_normal/  ##male normal spped 324 sentences (averaged speech-rate 6.4 mora / sec)
  02_fast/   ##male fast spped 324 sentences (averaged speech-rate 9.0 mora / sec)
  03_slow/   ##male fast spped 324 sentences (averaged speech-rate 5.0 mora / sec)
 -------------------------------------------------------------------------------------------