Abstract: An audio rendering device enhances voice audio such that audible voice is not overwhelmed by other aspects of the soundtrack. The device attenuates right and left channels in an audio stream in response to a detected voice component in the audio stream, and boosts the voice component in the audio stream based on the level of attenuation of the right and left channels. Voice components are distinguished from the non-voice components by separating center channel and mono information from the left, right and surround channels. Non-voice components are attenuated down towards a non-voice threshold level based on an attenuation ratio. Voice components are boosted up toward a voice threshold level, so that the spoken voice is more audible to viewers and not overwhelmed or drowned out by the non-voice aspects of the soundtrack.