What if I want to skip the stt and directly feed the voice to llm. I do not see this option