Comment by pyryt

Knowing when to speak is actually a prediction task in itself. See eg https://arxiv.org/abs/2010.10874

Would be indeed great to get something like this integrated with whisper, LLM and TTS