when is audio normalization needed

[codemaster]

If I capture some video footage, and the audio was captured by the microphone that's built into the camcorder, the volume seems to be already normalized (somehow automatically by the camcorder, maybe it's called AGC (auto gain control)).

Then, I edit this footage in Vegas, and then I use MeGui to compress it. Many compressors, like megui, staxrip, squeeze, etc, include a normalization option, to normalize the audio before compressing it.

But I'm not sure if I should enable normalization. I mean, for the audio captured with this camcorder, it seems redundant to normalize audio after editing in vegas and before compressing it to aac. Or, would you recommend that normalization should always be enabled in these compressors, regardless of what type of gain control was used to capture the audio ?

[hello_hello]

Normalisation (the type you can enable in conversion programs such as MeGUI and Staxrip etc) increases the over-all level so the volume peaks within the audio are at maximum. Often it'll increase the volume to some extent, but the whole audio has it's volume adjusted by the same amount. If the peaks are already at maximum, it'll have no effect.

If you're downmixing multichannel audio to stereo with MeGUI it probably pays to use it, as the formula MeGUI uses for downmixing has a tendency to reduce the volume, but when re-encoding stereo audio it's really personal preference. If you think the audio level is a bit low then there's probably no reason not to use it (at worst it'll have no effect if the peaks are already at maximum), but otherwise it's up to you.

The only possible downside of normalisation is it can change the relative volume between different audio tracks. For example you might have two audio tracks which over-all have pretty much the same volume, but one has a loud section with peaks already at maximum level, while the other doesn't. For the first track the volume won't change, but for the second it'll be increased, so the two tracks mightn't sound like they have the same average volume any more.

[codemaster]

Ok, then I'll enable normalization in megui only for things like videos captured with a tv tuner or with replay video capture, and I'll disable it for videos edited in Vegas (so that the volume remains exactly at the levels that I've specified with the audio envelope).

[hello_hello]

Even when converting "normal" video I don't use normalisation much myself these days, due to the possible variation is can cause between relative volumes. Not that it's likely to be huge, but it can happen. For movie sountracks it's no big deal (it's not hard to adjust the volume between movies) but for TV episodes I like to keep the relative volumes the same. Therefore I tend to convert "as-is" unless I'm converting multichannel audio to stereo, but even then I usually just downmix to stereo while applying a 6dB gain reduction to prevent clipping.

If you're really keen, there's a method for "normalising" audio called ReplayGain. It's designed for music but I've found it works pretty well with "soundtrack audio" too. The idea is it scans the audio using a formula for working out how loud it sounds to the human ear. It then saves the data as tags and the player should adjust the audio accordingly (it's hardly supported that way outside of PC audio players). Alternatively a conversion program can use the data when converting to another format in order to adjust the volume, rather than just use standard normalisation.

No video conversion GUIs I know of support ReplayGain. Some will let you make a manual volume adjustment while re-encoding but that's about it. As a result, if I want to convert audio while applying ReplayGain I use a different program. In my case, it's foobar2000 (in fact I use MeGUI for video conversion but mostly foobar2000 for audio). It'll scan the audio using ReplayGain, save the data in tags, and then you can create a conversion preset which applies ReplayGain while converting. If I'm re-encoding audio which has been "peak normalised" and the volumes are all over the place, I sometimes use ReplayGain as a method of trying to "un-normalise" them.

If you want to experiment, converting to MP3 while normalising with MeGUI might be one way. I suggest MP3 as it's the one format which can have it's volume adjusted losslessly without re-encoding. MP3Gain will scan MP3s using ReplayGain and adjust their volumes accordingly. It'll also do standard peak normalization (needs to be enabled in it's options). That way you could apply ReplayGain and/or reverse the volume changes MP3Gain applies nice and easily while comparing volumes with your ears to see if ReplayGain works for you. When converting "music" audio (CD tracks etc) I couldn't imagine not using it.