[hamsta]

The audio files from LibriVox are commonly used to train neural networks for ASR and TTS, see e.g. https://www.danielpovey.com/files/2015_icassp_librispeech.pd... and https://arxiv.org/abs/1904.02882