A method for communication includes establishing communication links over a packet network between a server and plurality of client computers that are to participate in a video teleconference. The server receives from the client computers uplink audio packets and uplink video packets, which respectively contain audio and video data captured by each of the client computers. The server mixes the audio data from the uplink audio packets so as to create respective streams of mixed audio data for transmission to the client computers and transmits to the client computers downlink audio packets containing the respective streams of mixed audio data. The server relays the video data to the client computers in downlink video packets. The client computers receive and synchronize the video data with the mixed audio data.