Audio models

Models for speech, transcription, and audio-native workflows.