A normalizer adjust the gain/volume so the highest peak in the audio waveform is below some threshold. It is the same constant gain there is applied (don’t change the dynamic range)
A compressor is reducing peaks above a threshold with the compression ratio and can apply makeup gain to increase the gain of the lowest peak (changes the dynamic range, so there is less difference between low and high peaks)
Both will change the perceived loudness.
The are many way to gain the same loudness level, but the result will sound different, your ears is the ultimate judge, but it is best to compare at the same loudness level, because if you listen to the same audio at 2 different loudness levels, you might think that the loudest one sound better, even if there is no difference.
I think I understand how a compressor works: everything above the threshold is compressed by the selected ratio. This reduces the dynamics of the sound, but also the volume. To compensate for this decrease in volume, you specify the value by which you want to increase the volume again in “Makeup gain”. But by how much? In “Gain Reduction” you can see approximately how much the compressor has reduced the volume by compressing. But you never know exactly how much you have to increase the volume.
In “Audio Loudness” you can read off quite exactly how loud the signal is, but you would have to watch the whole video carefully and you would also have to know which level should be how high.
Then I understood @Austin to say that this is very difficult and that’s why Normalize (Tow Pass) helps: This does not set the loudest peak, but the standardized LUFS, which I understand as a kind of “average loudness”. The prescribed or appropriate loudness can thus be set much better than if you look at “Audio Loudness” with your eyes and set it by hand.
Tim, I don’t mean to disagree with you, and so far I’ve gotten by without Normalize at all, but thought it would be a big help until I noticed that -15 LUFS sounds way too loud on me.
I was talking about Peak Normalization, the shotcut filter does loudness normalization, that looks at the RMS of the signal to find the constant gain to apply to the signal, even if can cause clipping of the signal.
Finally I understand the world again!
Hello @TimLau I couldn’t quite figure out from all that was said on this page about Normalize (Two pass) and also from what you write to me.
Now I know why: there seems to be a bug in the effect.
If I have already applied this filter and change the volume by Compressor EQ or whatever, then Normalize seems to go haywire. It makes some values that never seemed logical to me. Of course I have done analysis, but still totally illogical behavior. It doesn’t raise or lower the volume, as would be expected.
But:
But if I Normalize “Remove selected Filter” and then re-apply it, it does exactly what it is supposed to do: It then works correctly.
My Shotcut version is ARCH-22.01.30 on Linux.
It is too early to make a bug report, I need to investigate the phenomenon more.
Yes, the error occurs again and again. Also with shotcut-linux-x86_64-220306.AppImage.
Nromalize works properly only if it is applied last (“Add a Filter”). It does not matter where it is placed. “Move filter down” does not bring any improvement. You have to “Remove Normalize filter” and then select “Add Normalize filter” again. Then it works fine.