Statistical Consequences of Fat Beats

At the Audio Developer Conference ’21 in London, our founder Christian gave an insightful talk on the thinking and research that eventually lead to the development of Dynamic Grading.

Due to the dynamic nature of complex audio signals, objectively measuring and reasoning about properties like perceived loudness, dynamics or spectral balance is a surprisingly challenging task. At best, we might usually average some short-term measurement over a longer time (e.g. when measuring LUFS). In such a process, a lot of interesting information can get lost, especially for audio recordings outside the realm of modern mainstream music.

For practitioners in the creative process of music recording, production and mastering, judging dynamics and determining how to manipulate them (e.g. using a compressor) is among the most difficult things to learn. The process requires drudgingly acquired experience and critical listening skills, while visual aids are mostly limited to observing short-term cues such as level meters or real-time analyzers.

This talk introduces some new methods for visualizing, analyzing and manipulating audio, which allow for a more meaningful and intuitive assessment of what’s actually going on in practical music signals. Examples of music from different genres and epochs are shown and discussed, as well as single instrument and vocal tracks.

The talk ends with a quick presentation of an early prototype for Dynamic Grading. We’ve surely come a long way since then!