Speech API. Use /v1/audio/speech or /v1/audio/transcriptions.