Skip to main content
Topic: Handling Quiet Tracks (Read 285 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Handling Quiet Tracks

Hi!

I usually listen to recorded lectures, voices etc. that are not appropriately normalized/compressed so listening to these files is kind of impossible in a noisy environment. on Desktop I use https://www.foobar2000.org/components/view/foo_r128norm by @kode54 , which does quite a nice job.


  • Does Normalizing in general distort/alternate audio?
  • Is there some sort of Gain Module to louden tracks without compression/distortion?
  • Is there some sort of gain/compression module available for Mobile?

Re: Handling Quiet Tracks

Reply #1
Normalizing by definition won't distort anything within the file itself, it just means putting the loudest sample up to 0.0dBFS (or a user targeted level) and pushing the rest up the same.

Automatic Gain Control/Compensation algorithms all work differently, some will only attenuate, others will add gain, or limiting, or other types of compression, some will clip etc.

I'm pretty sure the ReplayGain built into FB2K will do what you want with the right settings, boost tracks to the max without clipping.

Re: Handling Quiet Tracks

Reply #2
Normalizing by definition won't distort anything within the file itself, it just means putting the loudest sample up to 0.0dBFS (or a user targeted level) and pushing the rest up the same.

Automatic Gain Control/Compensation algorithms all work differently, some will only attenuate, others will add gain, or limiting, or other types of compression, some will clip etc.

I'm pretty sure the ReplayGain built into FB2K will do what you want with the right settings, boost tracks to the max without clipping.

Thanks!
I tried it and works well if the track is not very Dynamic (e.g. <10 dB dynamic range)
but sometimes it does strange things. for example it attenuates a single track with +0.65 dB Peak Value, by -9.75 dB.

Is there some sort of meta-context based on which the gain value of per-file track gets calculated?


Re: Handling Quiet Tracks

Reply #3
This may happen when you are in album mode: f2k would apply gain to all files in the "album", maintaining their relative loudness.
Make sure you use the source mode = track. This should adjust gain individually to each track.


Re: Handling Quiet Tracks

Reply #4
This may happen when you are in album mode: f2k would apply gain to all files in the "album", maintaining their relative loudness.
Make sure you use the source mode = track. This should adjust gain individually to each track.


I used both:
-scan per-file track gain
-scan as a single album
and got exactly the same result (although I don't think there could any difference if you are selecting only 1 track).

Re: Handling Quiet Tracks

Reply #5
I used both:

You should not scan twice.

You might not have caught the basics. Scanning as album will also write track gain and peak tags.
There are four tags (here I assume FLAC for the field names):
<REPLAYGAIN_TRACK_GAIN> stores gain for the track
<REPLAYGAIN_TRACK_PEAK> stores peak for the track
<REPLAYGAIN_ALBUM_GAIN>: if a selection is identified as an album (for example when told to scan as an album), then this is a separate tag that is written alongside the track gain/peak tags.
<REPLAYGAIN_ALBUM_PEAK>: the peak for the entire selection.
Scanning "per-file" writes the first two. Scanning as album writes all four.

Then for playback, you go to Preferences -> Playback and select whether you want adjustment according to the "album" tags (defaulting to "track" tags if there are no album gain / peak written) - and whether you want fb2k to make sure you don't cross full volume.


But all this is distinct from on-the-fly normalization!

And, distinct from "applying" gain, which is a phrase one sometimes uses for changing the volume of the audio in the file, not just the tag.
Memento: this is Hydrogenaudio. Do not assume good faith.

 

Re: Handling Quiet Tracks

Reply #6
If you want to "compress" them, you'd have to re-encode, but the Dynamic Audio Normaliser would probably do a good job. I use it for compressing soundtrack audio for night viewing. The stand-alone version isn't particularly GUI firendly, but it's built into ffmpeg these days, so I use it that way.

If you create an encoder preset using ffmpeg as the encoder, this is an example of the command line I use for soundtrack audio. ffmpeg will convert to VBR MP3.

-i - -ignore_length true -af dynaudnorm=f=150 -c:a libmp3lame -q:a 2 %d

For converting to AAC I pipe from ffmpeg to QAAC. For the encoder preset the encoder is cmd.exe and the command line looks like this. I'm running XP but I believe it works on newer Windows versions.

 /d /c c:\progra~1\foobar2000\encoders\ffmpeg.exe -i - -ignore_length true -c:a pcm_f32le -af dynaudnorm=f=150 -f wav - | c:\progra~1\foobar2000\encoders\QAAC\qaac.exe --ignorelength -s --no-optimize --no-delay -V 91 -o %d -

The two options you might want to fiddle with are "f" and "g". Or don't specify either and try the defaults. Specifying the defaults would look like this:

dynaudnorm=f=500:g=31

https://ffmpeg.org/ffmpeg-all.html#dynaudnorm

 
SimplePortal 1.0.0 RC1 © 2008-2018