Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[murray]

Ive started this thread as we touched on the subject in a recent thread on something else.....

I run a number of PCM shorts, trailers, clips off Youtube and the audio levels are all over the place, I have to constantly monitor them and have differences of + to - 5db to keep a comfortable level.

One of the members hear told me to use the Analysis function on the clips and engage Volume Leveling and Adaptive Volume. Ive spent hours testing this since yesterday but still the levels are all over the place....It seems to work perfectly for just ripped CDs, but the video content is still + or - 5db to get the video clips similar in volume.

Is anyone else using this function on video content with DSP JRSS and does it work for you?

It would make my shows if I could get it to work as it does for ripped CDs...

[mwillems]

Ive started this thread as we touched on the subject in a recent thread on something else.....

I run a number of PCM shorts, trailers, clips off Youtube and the audio levels are all over the place, I have to constantly monitor them and have differences of + to - 5db to keep a comfortable level.

One of the members hear told me to use the Analysis function on the clips and engage Volume Leveling and Adaptive Volume. Ive spent hours testing this since yesterday but still the levels are all over the place....It seems to work perfectly for just ripped CDs, but the video content is still + or - 5db to get the video clips similar in volume.

Is anyone else using this function on video content with DSP JRSS and does it work for you?

It would make my shows if I could get it to work as it does for ripped CDs...

Try just using volume leveling instead of using it with adaptive volume.  Unless all the clips are loaded in a playlist together, adaptive volume will prevent the volume from actually being level.  If you want perfectly level volume all the time, just volume leveling is the answer.

[mattkhan]

You might find http://yabb.jriver.com/interact/index.php?topic=95807.5 gives a useful summary of what the different options do here.

[blgentry]

I've just done a few experiments and now I'm really confused about Volume Leveling with video.  This was done with MC 22.0.13 for Mac.

For starters, I tested Volume Leveling with Audio and it works as expected.  A fixed offset applied to each track.  I've been doing this for almost 2 years now, so I know it works very well.  Internal volume works with this, as you'd expect, turning the overall volume up and down, while keeping the volume offset.

Now a video track.  The Volume Level (R128) as analyzed is 6.3 LU, so it should offset the volume by 6.3 dB right?  It doesn't.  Instead, the offset is variable based on the Internal Volume Level.  Here are the results of the Volume Leveling offset, as it changes with Internal Volume:

Internal Volume     Volume Leveling Offset
0%                          +3.9 dB
5% - 80%               +13.3 dB
85%                       +11.4 dB
90%                       +8.9 dB
95%                       +6.4 dB
100%                     +3.9 dB

So 6.3 dB is no where to be seen and above 80% it seems to be maintaining a constant difference between the Internal Volume and the Offset, which yields essentially a constant TOTAL volume level above 80%.  (Every 5% in the linear range of the Internal Volume is 2.5 dB) Something seems very wrong with that.

This isn't how it's designed to work is it?

In case it matters:

Volume Protection:  Off
Loudness:  Off
Internal Volume Reference:  100

Brian.

[mwillems]

I've just done a few experiments and now I'm really confused about Volume Leveling with video.  This was done with MC 22.0.13 for Mac.

For starters, I tested Volume Leveling with Audio and it works as expected.  A fixed offset applied to each track.  I've been doing this for almost 2 years now, so I know it works very well.  Internal volume works with this, as you'd expect, turning the overall volume up and down, while keeping the volume offset.

Now a video track.  The Volume Level (R128) as analyzed is 6.3 LU, so it should offset the volume by 6.3 dB right?  It doesn't.  Instead, the offset is variable based on the Internal Volume Level.  Here are the results of the Volume Leveling offset, as it changes with Internal Volume:

Internal Volume     Volume Leveling Offset
0%                          +3.9 dB
5% - 80%               +13.3 dB
85%                       +11.4 dB
90%                       +8.9 dB
95%                       +6.4 dB
100%                     +3.9 dB

So 6.3 dB is no where to be seen and above 80% it seems to be maintaining a constant difference between the Internal Volume and the Offset, which yields essentially a constant TOTAL volume level above 80%.  (Every 5% in the linear range of the Internal Volume is 2.5 dB) Something seems very wrong with that.

This isn't how it's designed to work is it?

In case it matters:

Volume Protection:  Off
Loudness:  Off
Internal Volume Reference:  100

Brian.

Brian three thoughts:

1) Volume leveling works differently with positive offsets than negative offsets.  Only so much gain can be added at 100% volume before clipping occurs, and volume leveling is very conservative.  I note you didn't post the True Peak of the track, but if it's close to 0dB, then almost no volume could be added at 100% volume without causing clipping.  For example, if the tracks True Peak is at -2dBFS, then the most that volume leveling would try to add back at 100% would be 1dB and change (it always leaves a 1dB-ish margin of safety). As a result to get the full benefit of volume leveling with media that has a positive R128 value, you need to use an internal volume setting that will move the true peak far enough below 100% to give enough room (I have an 80% max volume setting for that reason among others).  The "scaling up" you're seeing is caused by JRiver trying to avoid clipping, and using more headroom as it becomes available. 

This works the same with Audio BTW, but you're much less likely to see the issue there because the vast majority of audio (for example, more than 99% of the 60,000 tracks in my collection) is engineered in such a way that it requires a negative volume leveling offset.  So unless you happened to be looking at the audio path with one of those one-in-a-hundred tracks with a positive offset, you'd never notice that audio works the same way. 

As to why the total doesn't agree with the analyzed value:

2) Any up or downmixing changes the calculus dramatically.  If, for example, you're playing a video with a 5.1 audio track downmixed to stereo, I'd expect to see a different number than the assessed volume leveling offset, because the downmixer attenuates the total significantly to avoid clipping due to inter-channel summing.  The attenuation attendant upon downmixing is not captured by the reported volume leveling offset (AFAIK), but a little while ago, Hendrik implemented some additional logic to try and "add back" some of the volume losses neccessarily incurred during downmixing (as part of volume leveling) when it's safe to do so.  So all bets are off there, and I would expect to see the kind of results you're seeing with downmixing because, while the attenuation from downmixing isn't reflected in the volume leveling offset, the gain "recapture" is.

3) Album vs. Track leveling.  Volume leveling exhibits different behavior with audio when all tracks are part of an album and when various tracks are in a playlist from various albums.  In the latter case each track gets it's own separate offset, but in the former case all tracks get the same offset, but that will not necessarily agree with any one of their volume leveling values.  So if you have multiple videos in a playlist, you may get different results depending on whether volume leveling is using the album or track logic for videos (I haven't done any testing there).

[murray]

I've just done a few experiments and now I'm really confused about Volume Leveling with video.  This was done with MC 22.0.13 for Mac.

For starters, I tested Volume Leveling with Audio and it works as expected.  A fixed offset applied to each track.  I've been doing this for almost 2 years now, so I know it works very well.  Internal volume works with this, as you'd expect, turning the overall volume up and down, while keeping the volume offset.

Now a video track.  The Volume Level (R128) as analyzed is 6.3 LU, so it should offset the volume by 6.3 dB right?  It doesn't.  Instead, the offset is variable based on the Internal Volume Level.  Here are the results of the Volume Leveling offset, as it changes with Internal Volume:

Internal Volume     Volume Leveling Offset
0%                          +3.9 dB
5% - 80%               +13.3 dB
85%                       +11.4 dB
90%                       +8.9 dB
95%                       +6.4 dB
100%                     +3.9 dB

So 6.3 dB is no where to be seen and above 80% it seems to be maintaining a constant difference between the Internal Volume and the Offset, which yields essentially a constant TOTAL volume level above 80%.  (Every 5% in the linear range of the Internal Volume is 2.5 dB) Something seems very wrong with that.

This isn't how it's designed to work is it?

In case it matters:

Volume Protection:  Off
Loudness:  Off
Internal Volume Reference:  100

Brian.

Brian I'm starting to get very confused with this myself on video.
I can have up to 8 videos in my playlist, all are 2 channel PCM.
Its not the louder ones that are a problem its the ones that were recorded very low, many are on youtube.
These ones are never raised in volume to match the rest when using volume levelling and analysis....

[mattkhan]

Its not the louder ones that are a problem its the ones that were recorded very low, many are on youtube.

do you mean you're using live playback?

[murray]

do you mean you're using live playback?

I don't know what you mean by that statement.

I have 2 channels video shorts, trailers etc etc in my library, I then put them into playlists to play in the order I select. This is where the problem is, the levels are all over the place when using volume levelling and analysis.

[mattkhan]

I don't know what you mean by that statement.

I have 2 channels video shorts, trailers etc etc in my library, I then put them into playlists to play in the order I select. This is where the problem is, the levels are all over the place when using volume levelling and analysis.

playing a video in youtube (or some other external app) with audio routed through jriver

[murray]

playing a video in youtube (or some other external app) with audio routed through jriver

No not streaming.... anything that came from youtube was first converted first to a mp4. Everything is mp4, mov or mkv in the playlists....

[blgentry]

1) Volume leveling works differently with positive offsets than negative offsets.  Only so much gain can be added at 100% volume before clipping occurs, and volume leveling is very conservative.  I note you didn't post the True Peak of the track, but if it's close to 0dB, then almost no volume could be added at 100% volume without causing clipping.

No, that can't be it.  Volume leveling is *designed* to work with files with all volume levels, not just files that are recorded low.   That's why Volume Leveling reduces the volume of tracks with high average volume levels.  The whole system is designed to work, transparently, without clipping.

2) Any up or downmixing changes the calculus dramatically.  If, for example, you're playing a video with a 5.1 audio track downmixed to stereo, I'd expect to see a different number than the assessed volume leveling offset, because the downmixer attenuates the total significantly to avoid clipping due to inter-channel summing.

This makes a lot of sense actually.  I *am* using a 2 channel playback system for these experiments, so I *am* downmixing from 6 channels to 2.  So I did another experiment:  I set up Null Output as my sound device.  Then I re-ran the test.  This time, I did NOT see the downmix.  It said it was sending 6 channel audio to the null output device.  But I was quite surprised to see that the output level *still* changes between 85% and 100%.  The value from 5% to 80% is less now.  But it still changes above 80%.  This can not be how the system is designed to work.  It makes no sense.

3) Album vs. Track leveling.  Volume leveling exhibits different behavior with audio when all tracks are part of an album and when various tracks are in a playlist from various albums. 

I can buy this argument too.  Make sense, especially because none of my movie files have a value for "album", so maybe they are all considered part of the same album.  So I tested video files all by themselves in Playing Now.  No change.

Something has to be wrong.

Brian.

[murray]

Try just using volume leveling instead of using it with adaptive volume.  Unless all the clips are loaded in a playlist together, adaptive volume will prevent the volume from actually being level.  If you want perfectly level volume all the time, just volume leveling is the answer.

I tested it that way many times and all the video files are still all over the place, some low some high!

[mwillems]

No, that can't be it.  Volume leveling is *designed* to work with files with all volume levels, not just files that are recorded low.   That's why Volume Leveling reduces the volume of tracks with high average volume levels.  The whole system is designed to work, transparently, without clipping.

I can assure you from quite a bit of testing (done while writing the volume leveling wiki article and beta testing the new volume leveling system in MC21) that volume leveling works the way I described it with files that have a positive volume leveling offset.  You can easily test it by playing a file with both: 1) a positive volume leveling adjustment and 2) a true peak at or above 0dBFS; at 100% internal volumes, such a file will receive 0dB of boost (or might even get a reduction) from volume leveling. 

This isn't really a design choice, it's a necessary consequence of digital audio: you can't get higher than 0dBFS in a digital system, so unless JRiver redefined 100% internal volume as something lower than 0dBFS there's nowhere for the boost to go without clipping on the peaks.  You're correct that volume leveling is designed to work transparently without clipping, but sometimes the preventing clipping part means that you can't get perfectly level playback on all files when internal volume is maximized because some files need boost and there's no headroom to add it.

It used to be the case that JRiver never applied the volume leveling boost when the file had a near 0dBFS true peak because of the risk of clipping.  I and others lobbied Hendrik to make it adaptable to when the internal volume was set lower so that it would scale up appropriately, and those files could achieve a level volume when there was enough headroom to accommodate them.  That way one could permanently set one's internal volume to 80% or 60% (or whatever) and get perfectly level volume all the time, even with tracks that couldn't be "leveled" at 100% volume.

More info here: http://yabb.jriver.com/interact/index.php?topic=99406.0

I promise that's the way the feature works, and that's almost certainly why you're seeing the scaling as you approach 100% volume.  If the file in question doesn't have a true peak that's closer to 0dBFS than the recommend volume leveling offset, then you might have found a bug.  If that's the case post the volume leveling adjustment and the true peak, and maybe we can get to the bottom of it.  Otherwise it's working as intended by scaling. 

This makes a lot of sense actually.  I *am* using a 2 channel playback system for these experiments, so I *am* downmixing from 6 channels to 2.  So I did another experiment:  I set up Null Output as my sound device.  Then I re-ran the test.  This time, I did NOT see the downmix.  It said it was sending 6 channel audio to the null output device.  But I was quite surprised to see that the output level *still* changes between 85% and 100%.  The value from 5% to 80% is less now.  But it still changes above 80%.  This can not be how the system is designed to work.  It makes no sense.

It makes sense if you listen to what I'm saying above.  You can't boost a 0dBFS signal without clipping.  A file can have a 0dBFS peak but nonetheless have a very low average volume level (such that R128 still recommends some amount of boost to reach the leveling target).  Such a track cannot be boosted at all without clipping if you have volume set to 100%. 

Similarly, a file that needs a +6dB boost to achieve a level volume, but has a true peak at -4dBFS, can only receive a maximum of 2dB worth of boost at 100% volume without pushing the peaks into clipping.  Volume leveling knows this and only applies what boost can be safely applied, but if you lower the internal volume, volume leveling recognizes this now, and will apply more of the needed boost until it reaches the correct value. 

The alternative is to never add the recommended boost when it could ever cause clipping (at any volume setting), which is a strictly inferior outcome as level volume is then unachievable (for those tracks) at any volume setting.

I tested it that way many times and all the video files are still all over the place, some low some high!

Can you post a little more info about your setup?  What kind of volume settings are you using?  Can you post the volume leveling adjustments and true peak levels for a few files that you think don't sound particularly level when played back to back?  If you reduce internal volume to 80% do you get level sound? 

These ones are never raised in volume to match the rest when using volume levelling and analysis....

For the reasons described above, the ones needing boost can't get it if you have internal volume all the way up, so if this is the case, you may just need to (paradoxically) turn down the internal volume to create the needed headroom to do all the leveling.

NOTE (post edited after initial post for clarity and to add examples)

[blgentry]

I can assure you from quite a bit testing (done while writing the volume leveling wiki article and beta testing the new volume leveling system in MC21) that volume leveling works the way I described it with files that have a positive volume leveling offset.  You can easily test it by playing files with a positive volume leveling adjustment and a true peak at or above 0dBFS; at 100% internal volumes, those files will receive 0dB of boost (or might even get a reduction) from volume leveling. 

You probably know that I know a little about digital audio and that I'm aware that 0 dB (fs) is the limit in digital systems.    So I understand the premise that boost above 0 dB doesn't make sense:  It will induce nasty sounding digital clipping.

I did your test with one of a few audio only files I have that qualifies.  It behaved exactly as you described:  It allowed boost when Internal Volume was set some number of notches lower than 0 dB, and eliminated the boost as it went up.

I haven't done enough experiments with video to be sure if volume leveling seems to work some of the time, most of the time, or mostly no at all.  I say this in terms of the listener experience; I'm not doubting that the functionality is working behind the scenes.  Rather, I'm doubting if it's something useful to human listeners with typical movie program content.  My gut feeling is that it doesn't work on most Movie and TV video content.

It makes sense if you listen to what I'm saying above.  You can't boost a 0dBFS signal without clipping.  A file can have a 0dBFS peak but nonetheless have a very low average volume level (such that R128 still recommends some amount of boost to reach the leveling target). 

...and this is why this *is* a design decision.  At some point, you have to decide how low you will make your volume leveling target.  You can not volume level everything because something recorded with an average level of -70 dB and a peak of 0 dB would need a volume leveling target of something like -60 to -69 dB (I'm not sure of the exact math, but I think you follow me).  That would obviously be very impractical as a volume leveling target for the entire system.  They had to compromise on a much higher volume leveling target.  I'm only troubled that video content (commercially produced movies) just don't seem to have any useful volume leveling to be done to them.

Brian.

[mwillems]

You probably know that I know a little about digital audio and that I'm aware that 0 dB (fs) is the limit in digital systems.    So I understand the premise that boost above 0 dB doesn't make sense:  It will induce nasty sounding digital clipping.

I figured you did, I just wanted to explain clearly in case we were talking past each other 

I did your test with one of a few audio only files I have that qualifies.  It behaved exactly as you described:  It allowed boost when Internal Volume was set some number of notches lower than 0 dB, and eliminated the boost as it went up.

I haven't done enough experiments with video to be sure if volume leveling seems to work some of the time, most of the time, or mostly no at all.  I say this in terms of the listener experience; I'm not doubting that the functionality is working behind the scenes.  Rather, I'm doubting if it's something useful to human listeners with typical movie program content.  My gut feeling is that it doesn't work on most Movie and TV video content.

As someone who has a system with the internal volume permanently set to 40% with volume leveling enabled, I can tell you that it works more or less perfectly with all analyzed video content for me.  I very rarely need to adjust the volume, and when I do, the needed change is very small.  There's a little bit of difference between native stereo audio, downmixed 5.1 audio, and downmixed 7.1 audio, but it's a few dB at most, and that's due to the downmixing recapture issue which is a little separate from the peak issue we're discussing here.

...and this is why this *is* a design decision.  At some point, you have to decide how low you will make your volume leveling target.  You can not volume level everything because something recorded with an average level of -70 dB and a peak of 0 dB would need a volume leveling target of something like -60 to -69 dB (I'm not sure of the exact math, but I think you follow me).  That would obviously be very impractical as a volume leveling target for the entire system.  They had to compromise on a much higher volume leveling target. 

They chose the cinema calibration target, which is, if I recall 83dB at -20dBFS.  You're right that no matter what target they chose, some hypothetical files wouldn't be able to be leveled appropriately at 100% volume.

I'm only troubled that video content (commercially produced movies) just don't seem to have any useful volume leveling to be done to them.

Commercially produced movies do get useful volume leveling if you don't set your internal volume to 100% (other than the downmixing recapture issue).  You just have to make sure you've left enough headroom for them, and turn up whatever volume control you have downstream from the computer a few dB (if you can).  What's the biggest differential you see in your files between a true peak and a R128 boost?  For me, setting internal volume to 80% or so would solve the issue for all of my films and TV shows; I have it set lower because my PC is hooked directly to a power amp, so 80% would be a little much most times.

[murray]

Did anyone come to any conclusion why the volume levelling doesn't really work on video content (2 channel)?
Have we concluded that it really only works on ripped CDs?

[mwillems]

Did anyone come to any conclusion why the volume levelling doesn't really work on video content (2 channel)?
Have we concluded that it really only works on ripped CDs?

Did you see the parts of my post addressed to you above?

Can you post a little more info about your setup?  What kind of volume settings are you using?  Can you post the volume leveling adjustments and true peak levels for a few files that you think don't sound particularly level when played back to back?  If you reduce internal volume to 80% do you get level sound? 

For the reasons described above, the ones needing boost can't get it if you have internal volume all the way up, so if this is the case, you may just need to (paradoxically) turn down the internal volume to create the needed headroom to do all the leveling.

Volume leveling works quite well for me for video content if I've lowered the internal volume enough to give it the necessary headroom, although there will still be some differences between 5.1 being downmixed to 2.0 and native 2.0 content due to the way downmixing works in JRiver. 

If you want to get to the bottom of it, try providing a bit more info about the files along the lines I mentioned.

[murray]

Did you see the parts of my post addressed to you above?
 

Volume leveling works quite well for me for video content if I've lowered the internal volume enough to give it the necessary headroom, although there will still be some differences between 5.1 being downmixed to 2.0 and native 2.0 content due to the way downmixing works in JRiver. 

If you want to get to the bottom of it, try providing a bit more info about the files along the lines I mentioned.

This is all I ever saw from you that you addressed to me. Everything else your were talking about was address to Brian, that's not me.

Try just using volume leveling instead of using it with adaptive volume.  Unless all the clips are loaded in a playlist together, adaptive volume will prevent the volume from actually being level.  If you want perfectly level volume all the time, just volume leveling is the answer.

I mentioned earlier I had done this and it never worked.

I don't care if its different for 5.1 or 7.1 material, I just would be happy with 2 channel levelling working as that's what I find most troublesome. Im having to raise volume on some 2 channel files up to 5/7db to bring it into an acceptable level. I spent days testing the analysis and volume levelling but Im still manually having to adjust it as I do when its all turned off. Since it doesn't work for me Ive just turned it all off.

All the files seem to be ACC.

Im happy to operate without it working just manually adjusting it all the time, but if Im doing something wrong, I would love to know what.

I use a number of files from youtube that I convert to mp4, these volumes are all over the place!