Dynamic Score is a scale that’s used to determine consistency in program loudness. It yields low values for tracks that are very compressed and limited, and whose loudness is quite consistent over the duration of the track. It yields higher values for tracks that could be compressed and limited, but have variance in dynamics over the duration of the whole track.
To determine the appropriate amount of existing compression and limiting, crest factor is used. Crest factor, when measured across the full track's length, represents the ratio between peak and RMS levels, expressed in decibels. A pure sinusoidal wave has a crest factor of 3.01 dB. Music content whose crest factor is below 3 dB could be considered degraded, and could be interpreted as noise rather than technically appropriate musical content. However, there's no technical limitation preventing a crest factor below 3 dB for musical content.
In Dynamic Score, dynamic variance across the full track’s length is determined by subtracting (full track’s) standard RMS level measured in dBFS from RMS peak (highest RMS-value in a 50 ms measurement window), and is expressed in decibels. High values alone would tell that there’s lot of dynamics, but doesn’t alone give hints whether it’s intentional or not.
Dynamic Score is a compound metric that uses three basic audio metrics in the calculation.
Using these three measures in an equation gives us an idea of the overall dynamics of the track. Since the Dynamic Score is used to determine track dynamics rather than micro-dynamics, crest factor is square-rooted to emphasize dynamic variance over momentary peaks.
\(DS={\sqrt C}\times ({RMS_{pk}}-{RMS})\)
Dynamic score has no theoretical upper bound since the crest factor can be ∞ for Gaussian noise. Lowest limit is 0 DS for pure DC.
A track that’s not compressed nor limited will likely have a high crest factor, but could be low in dynamic variance depending on the content. Dynamic Score will give a high reading for this kind of track, suggesting a need for processing. On the other hand, a track that is peak-limited will have lower crest factor, but could have a high dynamic variance. This track would also result in a high value. Tracks that have been aggressively processed have a very low crest factor, and very little variance. This track would result in a low Dynamic Score, suggesting no further processing is required for the track.
A modern metal song has RMS peak of -4 dB, and the overall RMS level of -8 dB. This is 4 dB dynamic variance. The track has however been mastered so that lots of dynamics have been sustained, thus it has a crest factor of 5 dB. The dynamic score for this track would be \(\sqrt 5 \times {(-4-(-8))}=8.94\).
The same track is then mastered again, but this time the crest factor is 3 dB. \(\sqrt 3 \times {(-4-(-8))}=6.93\)
This concludes crest factor alone doesn’t affect Dynamic Score linearly.
A jazz track has a RMS peak of -13 dB, and overall RMS level of -20 dB. It’s crest factor is 9 dB. \(\sqrt {9} \times (-13-(-20))=21\)
For comparison, if the same track was compressed so that the crest factor was 7, but dynamic variance went down to 4 dB, we see the relation of dynamic variance to Dynamic Score: \(\sqrt 7 \times (-13-(-17))=10.58\)
This concludes a change in dynamic variance is more important than crest factor alone.
Like LUFS, Dynamic Score is a compound metric. However, LUFS is designed to measure perceived loudness using psychoacoustic weighting (K-weighting based on equal loudness contours). For determining dynamic processing needs, this frequency weighting is unnecessary – only level variance matters. LUFS is also gain-dependent. A track measuring -14 LUFS, when attenuated by 9 dB, will measure -23 LUFS. Dynamic Score remains constant regardless of gain changes, as it measures ratios rather than absolute levels. This makes Dynamic Score more suitable for automated processing decisions that shouldn't depend on mastering level. From a computational perspective, LUFS calculation includes psychoacoustic filtering that's resource-intensive compared to the simple statistical measurements used for Dynamic Score. When processing large batches of audio, this efficiency difference becomes significant.
Dynamic Score provides a method for automated systems to assess dynamic range characteristics without relying on genre assumptions or target loudness standards. The assumption is that low crest factor combined with low variance indicates intentional processing choices, not accidents - content that has been deliberately compressed and limited to its intended dynamics.
Dynamic Score alone doesn't tell whether a track has been improperly processed, neither does it tell whether the track is up to loudness standards. It specifically addresses the challenge of determining processing intent: whether existing compression and limiting represents the content's intended character, or whether further processing would be appropriate.
Higher Dynamic Scores suggest content with either unprocessed dynamics or intentional dynamic range that could benefit from compression and limiting. Lower scores indicate content already processed to consistent loudness levels, where additional processing would likely degrade rather than improve the audio.
Dynamic Score is best used when the type of content is unknown - when an automated system must make processing decisions across varied material without manual intervention or genre-specific rules. Its gain-independence ensures consistent evaluation regardless of mastering level, while its computational efficiency makes it practical for batch processing operations.
Understanding what is loudness, how it's measured and why standards exist.