SpeechModel:
    | "best"
    | "nano"
    | "slam-1"
    | "universal"

The speech model to use for the transcription.