39. Japanese Expressive Speech Dataset Paired with Manga Images (MangaVox)
Data DOI
https://doi.org/10.32130/src.MangaVox
Producer, Project
National Institute of Advanced Industrial Science and Technology (AIST)
(Dr. Shinnosuke Takamichi, Dr. Tomohiko Nakamura and Dr. Hitoshi Suda)
Contents
MangaVox is a dataset for audio comics research that aligns manga images with acted speech recordings.
It contains recordings of dialogue, text outside speech bubbles, onomatopoeic and mimetic expressions, and other text appearing in eight manga works from Manga109. The recordings were performed by professional and semi-professional performers under the direction of an audio director. The dataset has a total duration of 14 hours, including 9 hours of speech.
Permission was obtained from the authors of the Manga109 works, and all performers consented to the public release of their recordings and their use for research.
SRC provides only the audio data; annotation data is available separately on the Creator's site.
Speaker
18 speakers (9 males and 9 females)
Recording environment
Studio
Speech File format
FLAC format (48kHz, 24bit, Mono)
Distribution media
1 DVD
Licensing
For non-commercial academic or technical research only
Price
No fee
Further information
Project Page: https://huggingface.co/spaces/imprt/mangavox