# voiceclap-extra-corpora

WebDataset shards of the additional voice corpora used in VoiceCLAP experiments, one folder per corpus: `voxceleb1/`, `voxceleb2/`, `ears/`, `expresso/`, `crema_d_emotion/`, `ravdess_emotion/`. VoxCeleb1/2, EARS and Expresso are the full-audio versions with FCaps-style captions (the smaller caption-only versions are in [laion/voiceclap-data](https://huggingface.co/datasets/laion/voiceclap-data)); CREMA-D and RAVDESS carry emotion labels. The licenses of the source corpora apply.