[Mac only] a speech-text foundation model for real time dialogue https://github.com/kyutai-labs/moshi