DynamicAudioNormalizer

[kolamorx]

Originally Posted by LoRd_MuldeR

If the original source was encoded in AC3/EAC3 (Dolby Digital) from, then you may try enabling the DRC (Dynamic Range Compression) feature of the AC3/EAC3 decoder already.

Apart from that, you can try the "--compress" option of the DynamicAudioNormalizer. See the manual for details...

Actually, my idea was to increase the volume before using dynaudnorm. (-af "volume=15dB,dynaudnorm=m=100...")
Is there a limit for the loss factor as there is for the gain factor?

[LoRd_MuldeR]

Originally Posted by kolamorx

Actually, my idea was to increase the volume before using dynaudnorm. (-af "volume=15dB,dynaudnorm=m=100...")
Is there a limit for the loss factor as there is for the gain factor?

If the volume can be increased any further using a fixed gain factor without causing clipping in the "loud" sections (i.e. there is some unused "headroom" in the file), then you should be doing this, of course.

However, if you use a tool like DynamicAudioNormalizer, it usually means that "traditional" normalization doesn't work on your file, because the "loud" sections are at maximum signal level already and yet the "quiet" sections are still too silent.

After all, the DynamicAudioNormalizer is not about "increasing" the volume (you could achieve that by simply turning up your speakers). It is all about "equalizing" the volume of "quiet" and "loud" sections

[kolamorx]

So, I guess I don't really understand how this tool works.
I tought that if, for example, the loud parts are 90dB, the quiet parts are 65dB, and you set the target value to 80dB- it decreases the loud parts by 10dB and increases the quiet parts by 15dB.

[LoRd_MuldeR]

Originally Posted by kolamorx

So, I guess I don't really understand how this tool works.
I tought that if, for example, the loud parts are 90dB, the quiet parts are 65dB, and you set the target value to 80dB- it decreases the loud parts by 10dB and increases the quiet parts by 15dB.

Yes, when using the DynamicAudioNormalizer, any "frame" whose level is below the target level will be amplified to the target level, and any "frame" whose level is above the target level will be attenuated to the target level – with some trickery (Gaussian filter) to avoid fast fluctuation of the amplification/attenuation factors. In other words, the DynamicAudioNormalizer will "equalize" the volume of "quiet" and "loud" sections.

The more simple "volume" filter will just apply the same fixed amplification factor to all samples. So, if the "loud" sections of your file already are at maximum signal level, then the "volume" filter can not further increase the volume of your file, as this would unavoidably result in clipping (in the "loud" sections). In other words, with the "volume" filter, the "loud" sections may prevent further amplification of the "quiet" sections. The DynamicAudioNormalizer can help in this situation.

If, on the other hand, the "loud" sections of your file are not at maximum signal level yet, i.e. there is some "headroom" in your file, then the "volume" filter can increase the volume of your file – to the extent that brings the "loud" sections to the maximum signal level. But even in that case, the "volume" filter does not "equalize" the volume of "quiet" and "loud" sections! It's just like turning up the volume of your speakers a little more. There is no "dynamic" adjustment.

Still, if you actually do have some "headroom" in your file, I would suggest to apply the "volume" filter first, to get "traditional" normalization. And then, if still required, use DynamicAudioNormalizer to bring up the "quiet" section some more...

[kolamorx]

How do I check how much headroom I have in my file?

[LoRd_MuldeR]

Originally Posted by kolamorx

How do I check how much headroom I have in my file?

For example, with the "volumedetect" filter:
https://ffmpeg.org/ffmpeg-filters.html#volumedetect

(it is max_volume that you are interested in)

[kolamorx]

Thanks.
Any chance you will increase the maximum possible value of the maximum gain factor in the future?

[LoRd_MuldeR]

Originally Posted by kolamorx

Thanks.
Any chance you will increase the maximum possible value of the maximum gain factor in the future?

I don't think so. An amplification factor of 100× already is equivalent to +40 dB.

If the "quiet" sections really are that much quieter than the "loud" sections that even an amplification by 100× (+40 dB) doesn't suffice, it probably means that the "quiet" sections are already lost in the noise.

Again: Don't use DynAudNorm to eliminate the "headroom" in your file. Use a "traditional" normalizer for that purpose. Use the DynAudNorm to "equalize" the volume of "quiet" and "loud" sections...

[manolito]

Originally Posted by LoRd_MuldeR

Still, if you actually do have some "headroom" in your file, I would suggest to apply the "volume" filter first, to get "traditional" normalization. And then, if still required, use DynamicAudioNormalizer to bring up the "quiet" section some more...

Not so sure about this recommendation...

I downloaded a short random clip with "sane" audio levels and did a few tests:

The source had these levels:
Integrated Loudness: -19.83 LUFS
Sample Peak: -6.07 SPFS

After conversion using DynAudNorm with default settings:
Integrated Loudness: -12.72 LUFS
Sample Peak: -0.45 SPFS

This time using peak normalizing first followed by DynAudNorm with default settings:
Integrated Loudness: -12.64 LUFS
Sample Peak: -0.36 SPFS


The loudness values are almost identical for both conversions. At least for "sane" source audio levels it is not necessary to apply regular peak normalization before DynAudNorm.


Just for fun I converted the same clip using the LoudNorm filter which is part of current FFmpeg versions. Since I do not like the defaults all that much I used the following parameters:

SET LUFS=-18.00
REM Target Loudness in LUFS
SET TruePeak=-1.5
REM Target True Peak value
SET L_Range=14
REM Target Loudness Range

I used 1-pass mode which of course has problems reaching the desired loudness values, but is much faster. The result came out like this:
Integrated Loudness: -16.27 LUFS
Sample Peak: -0.96 SPFS


To my ears this conversion did sound a little better than DynAudNorm, and it has the advantage that a target LUFS level can be specified.


Cheers
manolito

[kolamorx]

Originally Posted by LoRd_MuldeR

I don't think so. An amplification factor of 100× already is equivalent to +40 dB.

I mistakenly thought it's equivalent to +20 dB...

Originally Posted by manolito

To my ears this conversion did sound a little better than DynAudNorm, and it has the advantage that a target LUFS level can be specified.

I prefer dynaudnorm since it evens out the volume better.

[smok3]

Anyone using loudnorm and if so with what settings? Defaults are kinda ok, if a little sissy (lacks a bit of dynamics).

Have this in mpv input.conf

F1 cycle-values af "lavfi=[dynaudnorm=s=30]" "lavfi=[loudnorm]" ""

[manolito]

The effect of LoudNorm depends a lot on the parameters, it goes from subtle to very strong dynamic compression.

These are the FFmpeg defaults:

I, i
Set integrated loudness target. Range is -70.0 - -5.0. Default value is -24.0.

LRA, lra
Set loudness range target. Range is 1.0 - 20.0. Default value is 7.0.

TP, tp
Set maximum true peak. Range is -9.0 - +0.0. Default value is -2.0.

The target Integrated loudness is quite low, it complies to the ATSC standard. For EBU R128 it would be -23, for ReplayGain it is -18.

The target loudnesss range is small, it results in a strong dynamic compression.

True peak is OK, maybe it could be increased a little to -1.5 or even -1.


There is a GUI for LoudNorm called WinLoud (by Muxson). His defaults are optimized for podcasts, see them here:
http://www.muxson.com/winloud

The author of Loudnorm prefers different values again:
http://k.ylo.ph/2016/04/04/loudnorm.html


And myself I do not want too strong dynamic compression, and my target level is Replaygain. So I use

Target Integrated Loudness: -18 LUFS
Target Loudness Range: 14 dB
Target True Peak Level: -1.5 SPFS

Takes a little experimentation to find your preferred settings.


Cheers
manolito

[smok3]

Hmm, 7dB seems like it should be enough of dynamics, I guess I'll have to measure and compare that with something. p.s. I don't use anything-win.

[LoRd_MuldeR]

Originally Posted by manolito

The loudness values are almost identical for both conversions. At least for "sane" source audio levels it is not necessary to apply regular peak normalization before DynAudNorm.

Yes, with a properly produced/leveled recoding, the "loudest" peaks in the file should already be at maximum signal level, or just below the maximum signal level.

In that situation a "traditional" (non-dynamic) normalizer will do exactly nothing, because the "loudest" peaks prevent further amplification without clipping. But a tool like DynAudNorm (or maybe LoudNorm) might be used to bring up the "quiet" sections a little more – provided that there are some long enough "quiet" sections that are significant quieter than the "loudest" sections. Still, after such processing, a "traditional" (non-dynamic) normalizer will again do exactly nothing, because the "loudest" peaks still are at maximum signal level (or just below) and therefore further amplification is still prevented. Summary: With a properly produced/leveled recoding, a "traditional" normalizer is pretty pointless.

My point was: If you have a poorly leveled recording, where even the "loudest" peaks are way below the maximum signal level, i.e. the recording does not utilize the complete signal range, then the first step should be using a "traditional" (non-dynamic) normalizer in order to get rid of the unused "headroom". A tool like DynAudNorm is not needed (and probably not the best choice) for that purpose. If and only if, after the "traditional" (non-dynamic) normalization you still think that the volume of the "quiet" sections should be brought up more closely to the "loud" sections, then a tool like DynAudNorm is the right tool. Used in the right way, a max. amplification factor of 100× should be more than you'll ever need.

[hydra3333]

Originally Posted by smok3

Anyone using loudnorm and if so with what settings? Defaults are kinda ok, if a little sissy (lacks a bit of dynamics),

Not sure if this is what you mean ... I use loudnorm in ffmpeg to try to adjust recordings all to a usable level. Here's the code lifted directly from my script without all relevant variables defined, so buyer beware.

REM snip

set audiofreq=48000
set audiobitrate=384k
SET audiodelayadjms=010
set AudioDelayms=000
REM audiodelayadjms delays audio just a tad extra .01s so that it comes out after the lips move
REM AudioDelayms    is the final calculated delay for audio with audiodelayadjms added to it
SET lI=-16
SET lTP=0.0
SET lLRA=11

REM snip

REM to adjust Audio volume. find the loudness parameters in a first pass
ECHO !DATE! !TIME! ***************************** start Find Audio Loudness ***************************** >> "%vrdlog%" 2>&1
ECHO file "%~f1" ...
ECHO "%ffmpegexex64%" -threads 0 -nostats -nostdin -y -hide_banner -threads 0 -i "%~1" -vn -threads 0 -af loudnorm=I=%lI%:TP=%lTP%:LRA=%lLRA%:print_format=json -threads 0 -f null - >> "%vrdlog%" 2>&1
"%ffmpegexex64%" -threads 0 -nostats -nostdin -y -hide_banner -threads 0 -i "%~1" -vn -threads 0 -af loudnorm=I=%lI%:TP=%lTP%:LRA=%lLRA%:print_format=json -threads 0 -f null - 2> "%jsonFile%"  
SET EL=!ERRORLEVEL!
IF /I "!EL!" NEQ "0" (
   Echo *********  Error !EL! was found >> "%vrdlog%" 2>&1
   Echo *********  Error !EL! was found >> "%vrdlog%" 2>&1
   Echo *********  Error !EL! was found 
   Echo *********  ABORTING ... >> "%vrdlog%" 2>&1
   %xpause%
   EXIT !EL!
)
REM all the trickery below is simply to remove quotes and tabs and spaces from the json single-level response
set input_i=
set input_tp=
set input_lra=
set input_thresh=
set target_offset=
for /f "tokens=1,2 delims=:, " %%a in (' find ":" ^< "%jsonFile%" ') do (
   set "var="
   for %%c in (%%~a) do set "var=!var!,%%~c"
   set var=!var:~1!
   set "val="
   for %%d in (%%~b) do set "val=!val!,%%~d"
   set val=!val:~1!
REM   echo .!var!.=.!val!.
   IF "!var!" == "input_i"         set !var!=!val!
   IF "!var!" == "input_tp"        set !var!=!val!
   IF "!var!" == "input_lra"       set !var!=!val!
   IF "!var!" == "input_thresh"    set !var!=!val!
   IF "!var!" == "target_offset"   set !var!=!val!
)
echo input_i=%input_i% >> "%vrdlog%" 2>&1
echo input_tp=%input_tp% >> "%vrdlog%" 2>&1
echo input_lra=%input_lra% >> "%vrdlog%" 2>&1
echo input_thresh=%input_thresh% >> "%vrdlog%" 2>&1
echo target_offset=%target_offset% >> "%vrdlog%" 2>&1
REM check for bad loudnorm values ... if baddies found, use dynaudnorm instead
set AUDprocess=loudnorm
IF /I "%input_i%" == "inf" (set AUDprocess=dynaudnorm)
IF /I "%input_i%" == "-inf" (set AUDprocess=dynaudnorm)
IF /I "%input_tp%" == "inf" (set AUDprocess=dynaudnorm)
IF /I "%input_tp%" == "-inf" (set AUDprocess=dynaudnorm)
IF /I "%input_lra%" == "inf" (set AUDprocess=dynaudnorm)
IF /I "%input_lra%" == "-inf" (set AUDprocess=dynaudnorm)
IF /I "%input_thresh%" == "inf" (set AUDprocess=dynaudnorm)
IF /I "%input_thresh%" == "-inf" (set AUDprocess=dynaudnorm)
IF /I "%target_offset%" == "inf" (set AUDprocess=dynaudnorm)
IF /I "%target_offset%" == "-inf" (set AUDprocess=dynaudnorm)
REM
REM later, in a second encoding pass we MUST down-convert from 192k (loadnorm upsamples it to 192k whis is way way too high ... use  -ar 48k or -ar 48000
REM
IF /I "%AUDprocess%" == "loudnorm" (
   ECHO "Proceeding with normal LOUDNORM audio normalisation ..."  >> "%vrdlog%" 2>&1
   ECHO "Proceeding with normal LOUDNORM audio normalisation ..."
   set loudnormfilter=loudnorm=I=%lI%:TP=%lTP%:LRA=%lLRA%:measured_I=%input_i%:measured_LRA=%input_lra%:measured_TP=%input_tp%:measured_thresh=%input_thresh%:offset=%target_offset%:linear=true:print_format=summary
   echo "loudnormfilter=%loudnormfilter%" >> "%vrdlog%" 2>&1
) ELSE (
   ECHO "********* ERROR VALUES DETECTED FROM LOUDNORM - Doing UNUSUAL dynaudnorm audio normalisation instead ..." >> "%vrdlog%" 2>&1
   ECHO "********* ERROR VALUES DETECTED FROM LOUDNORM - Doing UNUSUAL dynaudnorm audio normalisation instead ..." >> "%vrdlog%" 2>&1
   ECHO "********* ERROR VALUES DETECTED FROM LOUDNORM - Doing UNUSUAL dynaudnorm audio normalisation instead ..."
   ECHO "********* ERROR VALUES DETECTED FROM LOUDNORM - Doing UNUSUAL dynaudnorm audio normalisation instead ..."
   set loudnormfilter=dynaudnorm
   echo "loudnormfilter=%loudnormfilter% (Should be dynaudnorm)" >> "%vrdlog%" 2>&1
)
ECHO !DATE! !TIME! ***************************** end Find Audio Loudness ***************************** >> "%vrdlog%" 2>&1
REM ------------------------------ audio conversion ------------------------------ 
echo !DATE! !TIME! ***************************** start ffmpeg %makeaudiotype% AUDIO conversion ***************************** >> "%vrdlog%" 2>&1
REM mp3 audio conversion "%~f1" ...
IF /I "%makeaudiotype%" == "mp3" (
ECHO  mp3 audio conversion "%~f1" ... >> "%vrdlog%" 2>&1
ECHO  mp3 audio conversion "%~f1" ...
ECHO "%ffmpegexex64%" -threads 0 -nostats -threads 0 -i "%~1" -threads 0 -vn -threads 0 -map_metadata -1 -af %loudnormfilter% -c:a libmp3lame -ab %audiobitrate% -ar %audiofreq% -threads 0 -y "%parmp3%"  >> "%vrdlog%" 2>&1
"%ffmpegexex64%" -threads 0 -nostats -threads 0 -i "%~1" -threads 0 -vn -threads 0 -map_metadata -1 -af %loudnormfilter% -c:a libmp3lame -ab %audiobitrate% -ar %audiofreq% -threads 0 -y "%parmp3%"  >> "%vrdlog%" 2>&1
SET EL=!ERRORLEVEL!
IF /I "!EL!" NEQ "0" (
   Echo *********  Error !EL! was found >> "%vrdlog%" 2>&1
   Echo *********  Error !EL! was found >> "%vrdlog%" 2>&1
   Echo *********  Error !EL! was found 
   Echo *********  ABORTING ... >> "%vrdlog%" 2>&1
   %xpause%
   EXIT !EL!
)
)
REM aac audio conversion "%~f1" ...
IF /I "%makeaudiotype%" == "aac" (
ECHO  aac audio conversion "%~f1" ... >> "%vrdlog%" 2>&1
ECHO  aac audio conversion "%~f1" ...
ECHO "%ffmpegexex64%" -threads 0 -nostats -threads 0 -i "%~1" -threads 0 -vn -threads 0 -map_metadata -1 -af %loudnormfilter% -c:a libfdk_aac -cutoff 18000 -ab %audiobitrate% -ar %audiofreq% -threads 0 -y "%paraac%"  >> "%vrdlog%" 2>&1
"%ffmpegexex64%" -threads 0 -nostats -threads 0 -i "%~1" -threads 0 -vn -threads 0 -map_metadata -1 -af %loudnormfilter% -c:a libfdk_aac -cutoff 18000 -ab %audiobitrate% -ar %audiofreq% -threads 0 -y "%paraac%"  >> "%vrdlog%" 2>&1
SET EL=!ERRORLEVEL!
IF /I "!EL!" NEQ "0" (
   Echo *********  Error !EL! was found >> "%vrdlog%" 2>&1
   Echo *********  Error !EL! was found >> "%vrdlog%" 2>&1
   Echo *********  Error !EL! was found 
   Echo *********  ABORTING ... >> "%vrdlog%" 2>&1
   %xpause%
   EXIT !EL!
)
)
echo !DATE! !TIME! ***************************** end ffmpeg %makeaudiotype% AUDIO conversion ***************************** >> "%vrdlog%" 2>&1

REM snip

Sill reconsidering dynanorm as the default.

[smok3]

^ That any different than ˇ ?

mpv --af=lavfi=[loudnorm] file.mp4

edit: Oh I see, your LRA is higer than default.

To test:

af=lavfi=[loudnorm=LRA=10]

[osgZach]

Just a quick question, and an observation if you will... (this is a big thread)

Q: How concurrent is the ffmpeg filter with your source builds? In terms of any major improvements. I'm probably going to be writing myself a huge script of some sort, or maybe even a program, to automate batch processing of audio tracks from movies (I'm trying to automate the whole encoding chain really, video, subs, etc), so the less tools I have keep track of the better. Keeping it all inside ffmpeg would be preferable.

The observation: I think someone may have posted something similar to this a few pages back, but I've pretty much been up and down the race track trying out different methods of downmixing multichannel (5.1, 7.1) audio tracks to Stereo and avoiding the dynamic range problems associated with it. At first when I was playing with dynaud in ffmpeg, I wasn't noticing much difference, then I downloaded the CLI off Mulder's repo, and noticed a big improvement in reaching what I was trying to achieve.

The one major difference was when I used Mulder's CLI I was operating on a Stereo track generated from ffmpeg with -ac 2. while trying the filter in ffmpeg though I was feeding it the AC-3 source directly though (I did throw -ac 2 before the filter also). There is a distinct difference in my opinion. So I re-ran the Stereo track through ffmpeg and got pretty much the same result as using the separate CLI . I'm not an audio expert by any means, but I would say there is around a 1dB difference or more in loudness. Is there a reason for that?

It seems like for my workflow I need to generate a Stereo track as an intermediate file, and then re-filter that file to get my final output.
It goes pretty fast so I don't really care, just found it odd.

if needed I can probably find some time to trim a sample clip and provide an example package if that suits you.

[sneaker_ger]

Originally Posted by osgZach

(I did throw -ac 2 before the filter also)

Does this really ensure the down-mix comes before the audio filtering (e.g. dynaudnorm)? I guess not.

[LoRd_MuldeR]

Originally Posted by osgZach

Just a quick question, and an observation if you will... (this is a big thread)

Q: How concurrent is the ffmpeg filter with your source builds? In terms of any major improvements.

There haven't been any significant improvements to the "core" algorithm, only a few fixes here and there.

[kolamorx]

Hello.
I've read your documentation about the maximum gain factor, that ususally it's not recommended to increase the value above the default but sometimes it may be necessary to allow higher gain factors.
So if I understood correctly (and from my experience), when you have to allow higher gain factors, it's better to set it as less value as possible.
What free tool for Windows do you recommend for volume measurement, so I can get data about the dialogue volume levels in some scenes per movie (which sometimes are very low), in order to set the maximum gain factor as less value as possible?