We present a new method for attaching data to digital audio-visual (AV) documents. Our method is steganographic, i.e., a secondary data stream is integrated or embedded into a given host data stream. Using a psychophysicmodel evaluated on the AV data we perform embedding in a perceptually transparent way. This is, the perceptual quality of the AV document is not altered. An important characteristic of our technique is the obtainable high embedding data capacity. Applications using this technique frequently embed content related information. As one of our examples, we describe a psychoacousticaudio codec for multiple generations compression.