Moshi is a speech-text foundation model for real-time full-duplex spoken dialogue, featuring a...