Best program to normalize the audio in a video to a target db (Windows)

[Herchi]

I use Xmedia Recode because it allows you to analyze the maximum volume found and raise it to a new limit (volume correction). It also allows you to normalize it dynamically (Dynamic audio normalizer).

It is almost perfect: I have some movies where the overall sound is low, so the volume correction is very useful, but maybe I ask for too much when I want the low sounds to not be so low (without touching the high ones).

I will try to explain it with an example: imagine a video with low sound, I analyze it with Xmedia Recode and it is around 60db. I apply volume correction to 92db, but applying "Dynamic audio normalizer" I see that it does not reach 92db because it also flattens/reduces the high sounds and it would no longer reach 92db.

I do not want to extract the audio with a program, edit it and reintroduce it into the video.

[ProWo]

This is not correct. Audio normalization works in such a way that you reach a maximum of 0dB (anything above that would mean clipping). You can achieve this by simply raising the volume to max 0dB (standard normalization) or with dynamic normalization, in which the extremely quiet passages are raised significantly and the louder ones hardly at all, but here too the peak should not be above 0dB.

FFmpeg uses the DynAudNorm filter for this. You can use it in a command line or with a GUI, e.g. clever FFmpeg-GUI.

But you cannot avoid re-encoding the audio track.

[Herchi]

But you cannot avoid re-encoding the audio track.

I dont want to touch the video, only audio.

clever FFmpeg-GUI makes a mp3: Xmedia Recode create a video file (mp4) with the new audio with the target db.

[ProWo]

clever FFmpeg-GUI makes a mp3...

clever FFmpeg-GUI makes any audio.

[Herchi]

clever FFmpeg-GUI makes a mp3...

clever FFmpeg-GUI makes any audio.

I want to rip a video only rendering the audio track

[ProWo]

If you don't like the sound in a movie, you can extract the audio track, change it and add it back in.
With clever FFmpeg-GUI, there is no extraction;
you just encode the audio track to a new file, which is then muxed to the original video without the old audio track.
Watch the video.


[Attachment 84748 - Click to enlarge]

[pandy]

Audio normalization works in such a way that you reach a maximum of 0dB (anything above that would mean clipping). You can achieve this by simply raising the volume to max 0dB (standard normalization) or with dynamic normalization, in which the extremely quiet passages are raised significantly and the louder ones hardly at all, but here too the peak should not be above 0dB.

two cents - firstly 0dB means nothing - dB is unit-less and need reference - i understand you are referring to 0dBFS (Full Scale) but for audio (or other small electric current related) engineer it may be 0dBm, 0dBV, 0dBmV etc, secondly 0dBFS level may and usually do clipping anyway - this is called inter-sample clipping (peaks) https://www.audioholics.com/audio-technologies/issues-with-0dbfs-levels-on-digital-aud...ayback-systems - technically blindly safe maximum level in digital audio is around -3.0103dBFS (i.e. maximum 0dBFS level reduced by 0.5 bit) so whenever blind (no measurement) audio level normalization is performed then 0dBFS is not recommended as maximum level but -3.0103dBFS (i.e. -6.0206/2) instead.

[Herchi]

If you don't like the sound in a movie, you can extract the audio track, change it and add it back in.
With clever FFmpeg-GUI, there is no extraction;
you just encode the audio track to a new file, which is then muxed to the original video without the old audio track.
Watch the video.


[Attachment 84748 - Click to enlarge]

camtasia: spectogram


xmedia and clever:



I tested two: in Clever could I set the maximum decibels to reach? as in Xmedia. I think no
I prefed Xmedia (It is more comfortable to use and I can set a target db).

[davexnet]

DO you mean this? In this example, It tells you the current volume is 82db and that setting normalize to 92db is a 10db gain.
This operation raises all sample levels the same amount, it's like a classic normailize.
But it doesn't warn you if you set it too high causing the highest peaks to clip.
This could be a problem

[Herchi]

DO you mean this? In this example, It tells you the current volume is 82db and that setting normalize to 92db is a 10db gain.
This operation raises all sample levels the same amount, it's like a classic normailize.
But it doesn't warn you if you set it too high causing the highest peaks to clip.
This could be a problem

yes, I mean this.

Recommended is 89db I set 92db. when I see it in Camtasia the picture I dont hear peaks.

I want two things (copying the video):
a) leveling to a target (92db)
b) normalize (after or before?) the volume

[davexnet]

What is this 92db? Is it some reference to the peak level?
Usually audio programs talk about 0db as maximum level before clipping.

[pandy]

Recommended is 89db I set 92db. when I see it in Camtasia the picture I dont hear peaks.

I want two things (copying the video):
a) leveling to a target (92db)
b) normalize (after or before?) the volume

There is no such thing as 89dB - this is typical disinformation based on some misunderstanding - 89dB is probably some average mean level (like RMS) but it is confusing - someone probably misunderstand LUFS and EBU R128 https://en.wikipedia.org/wiki/LUFS?useskin=vector and create something that mislead others.

[Cornucopia]

They probably meant 89 dBSPL RMS, but without those reference points, as well as some more (like pandy mentioned), it's basically meaningless.


Scott