I understand that, I was just making sure if this is their approach. Because for example, with deepgram, they have live transcription with built in VAD (Voice Activity Detection)