The People's Speech Dataset is among the world's largest English speech recognition corpus today that is licensed for academic and commercial usage under CC-BY-SA and CC-BY 4.0. It includes 30,000+ hours of transcribed speech in English languages with a diverse set of speakers. This open dataset is large enough to train speech-to-text systems and crucially is available with a permissive license.
Supported Tasks… See the full description on the dataset page: https://huggingface.co/datasets/MLCommons/peoples_speech_v1.0.
