It will be nice to have component where I can upload sound of washing machine for example, when recognize it, and trigger when washing machine stoped.
Continuously recording and processing sound like that would take a lot of resources. Especially compared to solutions that already exist for your proposed use case, such as vibration or power sensors.
But sound is much light when video.