receiving, at an electronic device, first audio data representing an utterance, the first audio data being received from an initiating device;

generating first text data representing the first audio data by executing speech-to-text processing to the first audio data;

determining, using natural language understanding functionality, that the utterance is a request to establish a communications session with a named contact;

determining that a user account is associated with the initiating device;

determining that a recipient device is associated with a contact user account of the named contact;

determining, prior to establishing the communications session, that the user account is pre-authorized to establish a communications session with the contact user account of the named contact;

after determining the communications session has been pre-authorized, establishing a first communications session for the initiating device and the recipient device, the first communications session to be an audio call using a real time communication protocol;

receiving first data indicating that first sounds were received by a first microphone of the recipient device during a first amount of time after the first communications session is initiated;

determining from the first data a first time that the first sounds were received;

determining that the first sounds correspond to detection of a first speech activity;

after determining that the first sounds correspond to the detection of the first speech activity, determining the first sounds are non-device directed audio;

receiving second data indicating that second sounds were received by the first microphone during a second amount of time after an end of the first amount of time;

determining from the second data a second time that the second sounds were received;

determining that the second sounds correspond to a second detection of speech activity;

after determining the detection that the second sounds correspond to the second detection of speech activity, determining the second sounds are non-device directed audio;

determining that a third amount of time between the first time and the second time is greater than a predefined temporal threshold value;

determining that the first communications session is to end based, at least in part, on the first sounds and the second sounds corresponding to the non-device directed audio received at the recipient device and the third amount of time being greater than the predefined temporal threshold value; and

causing the first communications session to end.