ESPHome corrects a conversation loop in which a voice satellite starts listening before its own spoken reply has begun. The satellite could transcribe that reply as a new request and answer itself, rather than waiting for the person to speak.
Satellites using ESPHome’s media player wait up to fifteen seconds for speech to start and keep the microphone closed while the response plays. If playback never starts, they skip the follow-up turn. Tests on Home Assistant Voice Preview Edition covered short and long replies; neither was recorded as a fresh request with the correction. This is psewar’s first ESPHome contribution.