Thanks. Yes, that's exactly what I want to do: raise the overall level (if it doesn't already peak out at -1dB), and also also reduce the dynamic range, especially where there are big differences.. It looks like I should use both a Compressor and Normalize. I was trying to Normalize first, but I can see why it's better to apply the Compressor first, and then let Normalize set the final maximum peak level. I'll try using the Compressor in the Audio Cleaning dialog since the one in the Mixer FX would only kick in after everything else (if I understand how the Mixer works).

@FunkyB

As Jeff tried tactfully to point out, it is not a good idea to try to do what you are doing. Batch processing, which is essentially what you are trying to do with your audio, (or video) only works well within a project when all the parameters of the recordings are the same. Mixing of takes from either different recorded sources or different times and places will mean possible fluctuating audio levels and tonal variations as all the clips will be subject to the same processing. In reality, each clip should ideally get it's own track and be individually processed through the mixer with the appropriate effects, leveling, and frequency curves applied before the final mixdown. Your ears are the best tool for the job if you know how to process each clip to get the audio to sound roughly the same as the surrounding clips. Beware of audio cleaners, they often take out just as much of what you want as they remove.

Hi,

Ditto what Jeff and CubeAce said.

Audio, like video, is a science, but adjusting it is a black art. There is no automatic adjustment that will do it right. I am definitely not an expert on this, so the others may jump in here.

I use VPX, so in addition to Normalize (maximum level), we have Normalize (EBU R128) which is a transceiver standard that normalizes the audio signal to -23 LUFS (loudness units relative to the digital full level). Unlike normalizing to the maximum level, this provides enough upward space for temporarily loud signals.

Both are quite different, as you'll see in my example below.

Example 1:

I have 3 consecutive video clips taken with my cell phone during a walk in the woods in winter. In the first clip, there is only the sound of my footsteps and breathing, but with far away noise of traffic on the highway. In second clip, there is the same but I call the dog and talk a bit. In the third clip, same, but I call the dog more loudly. There is some clipping in the second and third clips when I call the dog.

Normalize:

Whether I select all 3 clips and normalize at once or individually, I get +3.5 db, 0 and 0. This raises the first clip so now I hear my footsteps, breathing and background noise much louder in the first clip than the other two. This is not what I want.

Compression:

If I remove the normalization and then compress clips 2 and 3 to "Speech largely compressed," the background and some of the louder parts sound a bit tinny. If I now normalize all 3 together, the amount is the same as before. The main difference is that the fainter parts of my voice are now louder, closer to the loudest part.

The compressor in the cleaning interface is very rudimentary. A reasonable compressor would allow adjusting the compression ratio, compression threshold, attack time, release time, and makeup gain. You can see these in the compressor for a track in the mixer.

The compression threshold is when the compressor kicks in, that is, above this level, the lowest sounds will be closer to the loudest sounds, depending on the ratio. A higher ratio means more compression. What the compressor actually does, is reduce the output volume of the louder parts. See the compression image in this article. This is why makeup gain is needed, the higher the ratio, the more makeup gain required. Normalizing after applying the compressor will also bring the volume back up.

What to do:

I suggest putting the audio of all of the similar clips on the same track, nothing else on that track. Then do a mixdown to get a wave file; the single clip will be red. Right-click and open the External editor, likely Music Editor 3 unless you changed it in the Program Settings folders.

Now you can operate on everything together. You can split this object into pieces, but unfortunately, ME3 does not allow more than normalizing individual objects. As far as I know, any of the other tools affect the entire track. If you have Audio & Music Lab Premium or its updated version Sound Forge Audio Cleaning Lab (or ACL or VSCL), you can apply effects to individual (split) objects and/or to the project.

I suggest first normalizing, turning on the Limiter, and then applying other effects. The main compressor is under Mastering, Dynamics, and you now have good control to work with. Try to find the threshold point and try various compression ratios. Listen, adjust, listen, adjust, repeat... Apply any other effects.

Now listen to the track. There may be parts that are too loud, parts too soft, someone said something in the background that you want to bring up or get rid of, etc. You can do this part now or back in MEP. Turn on the volume curve, add points and raise or lower the volume where required. Exit and the red wave file will be updated. As I said, you can adjust the volume curve either in ME3 or in MEP. If you have background music on another track, you may want to adjust the volume curve in MEP to suit.

Alternatively, you can use the track tools from FX in the Mixer.

There is a lot more that you can do, of course, but a lot is iterative and you need to use your ears.

Difference with EBU normalization:

In VPX, applying EBU normalization to the 3 clips gives me volume levels of +10.8 db, -6.3, and -7.9, respectively. Now the background in the first clip is way too loud, so I have to adjust it down. It would be better to do a mixdown and then apply EBU normalization, I think.

Example 2:

When I do a lot of voiceovers or narration, my voice sometimes changes from take to take and day to day, and the dog barked. Some takes are quite short, others long. As a result, when I normalize, the normalizer has a problem with short takes and they come out much louder than longer takes. I have to adjust the volume manually, or I do a mixdown, then normalize, then adjust the volume with the curve either in VPX or in the external editor. I try to remove breathes by reducing the volume locally using the curve or by other means. I also use an equalizer or dynamic equalizer. There is much more, of course. The danger is always in making the sound worse.

There is no set answer, no automatic control that does it all; it's a black art.

John CB

Wow. Thanks for the detailed examples and advice. The situations described by browj2 are very similar to what I am dealing with.

I can see why there is no magic shortcut to fix all audio. I'll have to play around with it more, obviously, but I have a better idea of how the different tools fit together now. I've had pretty good results using the compressor setting "Speech Largely Compressed" for clips that need compression, so I'll probably start with that and then either normalize or adjust each clip volume manually. If there are a few points that are way too loud, I'll adjust the envelope to bring them down. You mention turning on the Limiter. Is that the Limiter in the Mastering Suite (simply On or Off)? That's the only one I could find mentioned in the manual.

@FunkyB

You mention turning on the Limiter. Is that the Limiter in the Mastering Suite (simply On or Off)? That's the only one I could find mentioned in the manual.

No, the Limiter that I mentioned is under the part about using Music Editor 3. However, I usually turn on the Limiter in MEP/VPX as there may be something that I missed and I don't want it to exceed the limit.

Here is what the SF Audio Cleaning Lab manual says about the Limiter, which is probably the same text as for ME3, " The limiter prevents clipping and the auto function enables the volume to be automatically optimized.  "

Here is the text about the limiter in the Mastering Suite of VPX, "The limiter prevents clipping by automatically lowering the level if it is too high. Quiet parts remain unaffected. In contrast to the compressor it tries to leave the basic sound as unaffected as possible. "

Here is an image from the book "Working with Audio" by Stanley R. Alton (highly recommended) showing what both the compressor and limiter do:

You can see that for the compressor, the higher sounds are lowered and the lower sounds are raised, starting from the threshold. I don't know what threshold the compressor in Audio Cleaning uses, which is why I suggest using the compressor in Music Editor (ME3) and adjusting the threshold so that fainter background noise does not get compressed. If you are satisfied with the results of the Compressor under Audio Cleaning, great. But you may want to do some tests to compare it with the Compressor in ME3.

In the lower image, you can see that a limiter cuts off everything that is above 0db, which may not be what you want. If the original is clipped, loud sounds distorted, then there is no information available (all bits are 1) and lowering the volume will not help. Thus, it's best to be careful to avoid clipping during recording.

Magix mentions that turning on the Limiter automatically lowers the level rather than cutting it off like that shown in the lower image. Thus, if there is distortion, it will probably still be there. But if there is no distortion and the sound level is just too high, using the Limiter will bring the volume down so that it doesn't distort, if I understand correctly.

John CB

@FunkyB

Hi

Have you tested a program called The Levelator (free) - it does work with Windows 10.

I have tested this on some seascape scenes I have with different volume levels and it has done a reasonable job on levelling out the differences in the audio levels.

From MEP I exported the movie as WAV file and dropped this on to Levellator - it returned a levelled and normalised file which can the be used to replace the audio in the project.

HTH

John EB

@johnebaker

Interesting. Thanks. I'll try it out.

@FunkyB

Just playing around with an animation in Xara. Here is what I think the Limiter in MEP probably does:

Do others agree?

John CB

@browj2

Hi John

That is close to the normal function of a limiter as I understand it - on a good limiter, there should be 2 settings to control the limiting action which should come in gradually once the signal reaches a threshold and then progressively limit to the set maximum value.

John EB