A method and apparatus are disclosed that enable an interactive voice response (IVR) system to deliver content streams of various media types (e.g., video, audio, etc.) to telecommunications terminals. The illustrative embodiment provides extensions to the Voice extensible Markup Language (VXML) standard that enable an IVR system script to specify the playback order, timing, and coordination of multiple content streams (e.g., whether an audio stream and a video stream should be played back concurrently or serially; whether a particular content stream should finish before playback of another content stream commences; whether a content stream that is currently playing should be stopped and supplanted with another content stream, etc.).