Do these devices actively record conversations 24/7 when activated with "wake words," or do they just process short audio buffers before the trigger phrase? How does the balance between local processing and cloud computation work in this case?
Do voice assistants really listen to us all the time?
👁️ 98 views💬 1 replies❤️ 0 likes
1 Replies
Bro, I've been using an Echo Dot and a Google Nest Mini together at home for two years now; both have their "wake-word" detection running locally (on the chip). So the mic is always on, but the audio is only kept in a few-second buffer and sent to the cloud when it picks up a trigger word like "Hey Alexa" or "Hey Google." Before the trigger word, the device doesn’t record anything—it’s just a low-power DSP (digital signal processor) circuit listening for efficiency.
If privacy is a concern, I’d recommend two practical steps: 1) Put a physical cover over the mic or use the mic-off button (if available) on the device; this stops even the local DSP from listening. 2) Disable the wake words like "Hey Alexa" and only activate the mic by pressing a button (push-to-talk) when you want to give a command. This way, your voice only goes to the cloud when you decide, and the device doesn’t listen at all otherwise. With these two steps, you control the local/hardware-level listening and minimize data flow to the cloud. Honestly, after trying this, my sense of privacy improved a lot.