SpeechModel: "best" | "nano" | "conformer-2"

The speech model to use for the transcription.