I have an audio processing disorder and earmuff mode is a necessity for me to understand anyone in an instance with more than a few people. As-is, even with earmuff mode tuned as finely as I can tune it, I still end up saying "what?" a lot, unless I've cranked it down so far and gotten so close to someone that I can't hear anyone else, and when a big crowd ends up occupying a small, close-in space (as often happens naturally in VRchat), I get terrible sensory overload and have to leave, or at minimum take breaks from the conversation. There are a few issues I have with the earmuff system, and I think changes to them would help rectify these problems.

The directionality currently allowed is quite helpful, allowing me to focus my hearing on people I'm trying to listen to, but the minimum setting allowed is still fairly wide, which presents issues when many others are nearby.

The simplest fix for this would be to lower the minimum angle for the directionality slider.

A more complicated fix would be to add an optional, narrower cone in the middle of the directional cone, to allow me to specifically the voice of whoever I'm looking at.

Even better, and hardest to implement, would be to add user-based lock-on focus: if I look at someone who's talking, it amplifies their voice, and keeps doing so until I look at someone else who's talking. Integration with eye tracking would make this a seamless experience (for folks who've obtained or built eye tracking solutions).

Any of these options would make it much easier to understand my friends.