A multi-participant videoconference system incorporating a back-channel connection and a client video mixer is disclosed. The multi-participant videoconference system includes a client component and a server component. The server component provides a composite conference video signal to the client component. A region is defined in the composite conference video signal and the size and coordinates of the region are communicated to the client component by the server component over the back-channel. The client component captures local video and mixes local video into the composite conference video signal using the size and coordinates received from the server component for display.