Honestly, as someone who has been using this functionality almost daily for months now, the times that break immersion the most by far is when it does human-lik

Honestly, as someone who has been using this functionality almost daily for months now, the times that break immersion the most by far is when it does human-like things, such as clearing its throat, pandering, or attaching emotions to its responses. Very excited about faster response times, auto interrupt, cheaper api, and voice api — but the “emotional range” is actually disappointing to me. hopefully it doesn’t impact the default experience too much, or the memory features get good enough that I can stop it from trying to pretend to be a human