Many months ago I posted feedback on Adobe’s portal regarding what I refer to as a ghost loudness meter preset for Netflix measurement and compliance.
The Integrated Loudness submission target for Netflix referenced in “Near Field Audio Prerequisites for Mix Facilities” is -27 LKFS [+/- 2 LU tolerance] with True Peaks not exceeding -2 dBTP. They suggest setting an inserted limiter ceiling to -2.3 dBFS to “prevent false positives from minor differences in metering.”
Netflix has reverted to the original ITU algorithm (ver.1) that supports Dialogue Intelligence. They’ve adopted the use of Dolby’s Speech Gating Reference Code. This method of measurement relies on targeted speech to serve as the measurement anchor providing the measured dialogue bias is above 15%. Note: EBU R128 and ATSC A/85 specifications utilize a threshold sensitive Relative Level Gate measurement model.
Loudness Meters capable of measuring audio slated for Netflix deliverables include supplemental descriptors such as Integrated DIAL [the speech gated dialogue], DIAL LRA [as opposed to the program LRA], and in most cases a percentage descriptor indicating the dialogue bias vs. program material. FWIW they suggest an LRA of 10 LU or less for dialogue.
What is the basis for the -27 LKFS target?
Netflix engineers felt the ITU BS.1770-4 (2,3) algorithms utilizing a relative level gate over estimate the measurement of wide dynamic range audio. In essence highly dynamic audio may exhibit compromised dialogue intelligibility.
Engineers concluded speech/dialogue typically sits approx. -3 LU below all components in a -24 LKFS mix. Since consumers typically set playback level based on dialogue, -27 LKFS establishes optimal measurement translation.
Is Adobe Audition’s new meter compatible?
The meter includes a Netflix preset. As far as I can tell Adobe has not implemented the Dolby Reference Code necessary for Netflix compliance measurements. They’ve simply implemented a standard preset with a -27 LKFS Integrated Loudness target. Lack of support is [I believe] obvious due to the absence of the previously mentioned supplemental descriptors.
I’ve executed tests with discrete audio files passed through various loudness meters that feature dialogue gated measurement. The long term Integrated DIAL description was consistent across active meters. However this was not the case when using the Adobe meter with the Netflix preset selected.
FYI any meter with coded dialogue intelligence support may be customized. E.g. – you may customize the Integrated Loudness target to conform with your platform delivery requirements.
I use Audition daily. It’s a great tool. All good. Personally I do not produce mixes intended for Netflix. However Netflix compliance is now ubiquitous in professional post production. Supported tools and presets must exhibit proper implementation and support.