Implications of Object-Based Audio Personalisation Controls For Dialogue Intelligibility and Broadcast Loudness

Dialogue intelligibility is a persistent around of concern for broadcasters of audio-visual content, where dialogue often gets lost in a busy mix of varying audio elements, and audiences must either turn up the content to catch the dialogue or turn on subtitles. Given the current channel-based approach to mixing, broadcast content cannot be altered once a mix is finalised and broadcast, meaning the dialogue component in the mix cannot be independently increased in volume by the end-listener. This work investigates the concept of dialogue intelligibility in relation to broadcast loudness and the implications of object-based audio – a technology that enables greater audio personalisation controls over the soundtrack elements. Original test data is presented from user testing, recording the personalisation of average listeners of audio-visual content, in comparison with individuals with audio engineering backgrounds. The findings reveal that users on average set the level of the content 2.2 LUFS higher between the original and reduced-dialogue versions, emphasising the importance users place on dialogue for setting their base-level volume preferences. In the object-based phase, the audio engineer test group set the loudness higher on average by 2.8 LUFS, with the dialogue mixed 2.1 LUFS lower than the original. By contrast, the average listener group mixed the loudness very close to the original source material, with very similar loudness separation between dialogue and background content. The work concludes that the disparity between audio engineer and average listener loudness and mix preferences is a likely factor in creating inadequate mixes, where object-based audio may pose a solution to such situations.