Techniques are disclosed for managing media playback among an ad hoc group of playback devices. Such techniques may involve building a session among the playback devices in which playback devices communicate information regarding their playback capabilities. Based on the playback capabilities of the devices, playback mode may be derived for the session. Playback operations may be synchronized among the devices that are members of the session, in which devices receive identification of asset(s) to be rendered pursuant to the playback operation and timing information of playback of the asset. The devices may stream the playback assets directly from media sources when they are capable of doing so. In this manner, communication resources are conserved.