no but whatever deepgram transcribes is okay if the LLM will adjust or if the TTS is already trained to do that