What should a smart-glasses audio interface make clear?
It should make the active microphone, listening state, request progress and stop action understandable. The wearer needs to know whether to speak, wait, repeat or leave the interaction.
Feedback should explain the next action
Our recommendation is to keep states distinct. Listening means the system is accepting input; processing means it has moved on; disconnected means a required path is unavailable. A single indefinite animation hides those differences and can encourage repeated requests that make the situation worse.
If recognition fails, say that the request was not understood and offer a retry. If the remote service fails after recognition, preserve enough context to retry the operation without asking the person to repeat everything unnecessarily. The interface should reflect what happened rather than treating every failure as silence.