A communications system that supports multimedia components is easily adapted to existing network elements. Voice components arriving at or coming from a user having multimedia capabilities are referred from a telephony server serving the user to a multimedia server. A determination is made as to whether the other party supports multimedia capabilities. If that determination is negative, the component is passed back to the telephony server with an indication that the session is coming from the multimedia server to avoid an infinite loop. If the determination is positive, a parallel multimedia component is established between the parties while the multimedia server remains aware of the bearer path.