Audio filter "normalise two pass"

My setting: target loudness: -23.0 LUFS

My expectation was, when pressing “analyse”, this filter searches for the loudest point in the section and ensures that the loudest point does not become louder than -23 dBFS in audio peak meter in this section.

But I found that there is a point in which it is > -20 dBFS according to audio peak meter.

That’s confusing.

Thank you.

Actually, the target loudness refers to the whole audio clip where the “normalize two-pass” filter is applied.
Then you can see that the filter will try to make the integrated LUFS value (LUFS I) close to the target LUFS value after the clip is played back.
This will be displayed on the LUFS meter.

Here is an audio that is louder than -23 dBFs to which I apply the filter by setting LUFS to -23 dBFs.
The integrated loudness (I) finally matches the adjusted value (-23). At some moments the loudness is higher but this filter analyses the whole audio clip (the noisy parts and the quieter parts) and the final result is very close to the target LUFS value.

I found an explanation that may shed some light on these concepts:
What are LUFS? Loudness Metering Explained


Thank you @ejmillan .

I have made further experiments:
I have displayed LUFUS meter.
My setting: target loudness: -23.0 LUFS
I press analyse.

When analyse is done, I play this section.
The integrated LUFS value (LUFS I) goes up until a particular value. For example -24.

Then I go back to the beginning of the section.
I do not press the “reset” button.
Then I play the section again.
Now the integrated LUFS value (LUFS I) is different.
For example it goes now until -21 .

Does this mean, that in the second play of this section the sound is played louder, when I do not press the “reset” button?

If you do not press reset, the LUFS meter will start from previous data and show new measurements taking into account the previous data.
Then the integrated value is influenced by measurements prior to the reset.
The LUFS Meter is not like the spike meter. LUFs I is like an average value. If you do not reset the values obtained, they enter the calculation and return other values that include the audio previously recorded in the playback.
The built-in value changes during playback but eventually the value should get close to the target LUFS value. Since it is an average, low audio levels (such as fade-in or fade-out) are part of that average.
I did not perform the test but my hypothesis is that exporting that audio (normalized to -23 LUFs) will give you audio with that integrated value, regardless of whether the LUFS meter shows other values (for not restarting the meter)
Just do the test. Export a test audio, normalized to -23 LUFs (or whatever you need).
Open a new project in Shotcut, embedding that audio, and play it. The I value (at the end of the file playback) should be -23 LUFs (You analyzed all the audio, so the setting is for the entire length of the audio)

You can use an online LUF meter to compare the results.

This meter incorporates options that discard very quiet parts of the audio so the results may not match exactly. However, the audio file will always give the same loudness measured with the same tool.

This topic was automatically closed after 90 days. New replies are no longer allowed.