AudioModel: "whisper-1" | "gpt-4o-transcribe" | "gpt-4o-mini-transcribe"