Video conferencing with reduced bandwidth usage is disclosed. One method comprises receiving an audio stream at a server station connected to a plurality of participants in the video conference, at least one participant being a client station connected to the server station by a network and supplying the audio stream, the server station including a first memory and processor, and the client station including a second memory and processor. A mixing algorithm is performed that determines which video streams of the participants to mix to form a conference video stream for display by at least one of the participants. A video stream from the client station is selectively transmitted to the server station responsive to the mixing algorithm, and the server station transmits the conference video stream and/or the streams to be mixed to the participant(s).