Audio - Normalize One Pass vs Two Pass

[ken_wawa]

Hi!
What are the differences between one pass and two pass normalize filter from practical point of view?
Lately I’ve played a bit with them and I have impression that one pass gives slightly more nicer to ear result when set to the same target loudness. Moreover I have impression that two pass somehow highlights reflections and make the final result to sound a bit like audio registered in long tube. It might be somehow related with in material quality but still it’s interesting that in case of one pass filter it’s less visible (or rather it’s heard less).

[shotcut]

The main difference is that one pass must be dynamic/reactive. The first pass in a 2 pass scenario does analysis, and the second pass uses the results of the analysis in a static manner.

[ken_wawa]

So in theory they should give similar results as the main algorithm is the same but just one can give slightly louder or less loud audio (just because analysis is not so perfect)? However this doesn’t seem to explain why minor difference in loudness cause some “reflections effect”…

[shotcut]

They both only affect gain. I do not hear what you are describing.

[brian]

There is a description of the two filters here:

Introducing the One Pass Normalize Filter and Audio Loudness Scope

The previous audio normalize filter has been renamed to “Normalize: 2 Pass”. Its functionality has not changed. You can select a clip, analyze loudness of the clip and then the filter will apply a fixed gain “shift” to achieve the target loudness across the clip. The target loudness is configurable. This type of normalization is most appropriate for pre-mixed clips like music or clips where the volume is constant for the duration of the clip. A new audio normalize filter has been added named “N…

[Austin]

@ken_wawa

in theory they should give similar results

Only if the source audio has the same volume level from start to finish. Otherwise, the results could be very different.

The 2-Pass filter considers the average volume level of the entire audio clip, and raises or lowers the volume of the entire clip to hit the average loudness target. If the source audio has loud peaks and quiet valleys, those peaks and valleys will be preserved in the filtered audio. They will all be raised or lowered by the same amount, like turning a physical volume knob on a music player.

The 1-Pass filter works on a sliding window of time. Valleys may get raised and peaks may get lowered, which could bring the overall clip to a flatter sounding volume level.

The filter to choose depends on the result you want. For an unprocessed speech clip, a consistent volume level may be the top priority, meaning the 1-Pass filter is the better tool. But when importing professionally produced music that is intentional about its peaks and valleys, it may be better to preserve those peaks and valleys by using the 2-Pass filter.