On your phone and other “personal” devices it waits until you press a button and then it listens.
On a home automation or other “smart” device, there are two programs running at once, one program records a certain number of seconds on a loop, this recording is local only and runs all the time overwriting itself as it goes.
A second program looks at the first program while it is running and if it detects the “wake” word it sends the snippet of voice to the cloud api to translate into an action.
Your voice only leaves your device after it detects a wake word though this is not a perfect system and false positives occur.
*If your phone is listening without a button press, that is an accessibility feature you enabled and you should probably disable it to save on your battery even if you aren’t paranoid
Latest Answers