A device that is connected to a conference receives broadcast messages from other devices. The device selects at least a subset of the broadcast messages. Each broadcast message in the subset includes an indication that a respective device transmitting the broadcast message is connected to the conference. The device extracts respective audio synchronization values from the broadcast messages in the subset. The device identifies, based on the respective audio synchronization values, an audio frame of an audio stream of the conference. The device then outputs the audio frame.