A collection of audio recordings in multiple languages used to train speech recognition and synthesis systems.