Emilia: An Extensive, Multilingual, and Diverse Speech Dataset for Large-Scale Speech Generation
Paper • 2407.05361 • Published • 2
speech-latents-v1
Audio latents with transcripts, emotion captions and emotion/quality scores, derived from the Emilia speech dataset (amphion/Emilia-Dataset; He et al., 2024, arXiv:2407.05361). Emilia is licensed CC BY-NC 4.0; this derivative is released under the same license, for non-commercial use only.
WebDataset tar shards in data/. Each sample has <key>.npy (float16 latents, [frames, 128]),
optionally <key>.ref.npy, and <key>.json (text, language, speaker, duration, dnsmos,
emotion_caption, emotion_annotation).
- Downloads last month
- 19