← StayDemo · no key, no microphone
Ready when you are.
This plays a real recording of a barking dog through the same detector a live session uses. Nothing here is faked or on a timer — the strip is drawn from the actual signal, and the response is released by the same rule that runs when a microphone is listening.
Watch for the gap. Stay does not answer while the dog is barking. It waits until the barking stops, then speaks into the quiet — answering a bark would teach a dog that barking works.
Sound on. The voice you'll hear was made with ElevenLabs Voice Design; in a real session it's your own.
One number is changed here, and this is it: after speaking, Stay normally refuses to speak again for 90 seconds. This clip is 40 seconds long, so at the real setting you would see the loop once and nothing else. The demo uses 20 seconds instead, which fits two responses and one refusal into the clip. Every other rule — the wait for quiet especially — is untouched.