Hey and thank you! If System Audio Awareness or the MIC weren’t active, CielChan won’t be listening. In that case, it’s possible the AI was hallucinating from other context, like memory, current window/screen awareness, or past observations. She should only react to live audio when those audio settings are enabled in the app and allowed by your computer’s permission settings.
I’m currently getting v1.6.6 ready, which will include an improved model if you’re using the standard local option, along with better memory and context handling. That should help reduce edge cases like this.