Audio levels

[Lal Chakraborty]

Hi,

I shot an eighteen minute short film. I edited on Vegas movie studio. The video was ok but I had problems with the audio.

1. There were a few shots where we had audio problems (speech was too faint) and to fix it I normalized the audio. The background noise increased as well as the speech. I used magic audio cleaning tools on that and cleared some noise. But somehow I get a feeling that I shouldn't have used the normalize function. It changes the sound quality too much. Can you tell me if I should be using it or not.

2. The different clips are at different audio levels. Specially after I added music tracks and special audio tracks. How do I make the levels of all these shots the same? Right now the audio levels are very erratic.

Thanks.

Lal.

[Jerry Porter]

I don't edit in Vegas, but why not just turn up the gain on the tracks that are low? Then run it through the cleaner and add some vocal compression afterwards. Also the best way to prevent this problem is to constantly monitor your audio while recording. Headphones are your friend.

[Lal Chakraborty]

Is there any way to achieve that over the entire film. There are hundred or so clips. It is difficult to manually adjust each level and make sure that the levels all match up.

[Don Bloom]

Vegas has great audio toools to accomplish this with. FIRST take the highest acceptable levels and make them your benchmark. Now to get the low levels up to that you have any number of ways to go. You can use an audio envelope (rubber band) and set nodes or points where you need to raise the levels in ceratin parts. You can try using the normalization switch, you can use apply non-real time FX and use volume or amplitude or of course you can use a combination of the tools. Be aware that anytime you bring up the levels there is a possiblity of raising the floor or ambient noise but that can be controlled with a number of tools, such as graphic EQ or track eq applied to the clip. Another thing you can do is to split the clips according to the levels and run each of them as a group on seperate tracks then apply the FXs you need to that track only AND run the various tracks to their own bus and run the FX to that bus. You can also control levels ot the bus.

There are so many things you can try its really a matter of trying different thing until you get the sound you want.

Don

[Jon Fairhurst]

Take a look at post #23 in this thread:

http://www.dvinfo.net/conf/showthrea...t=72734&page=2

[Jarrod Whaley]

I hate to be Reverend Preachalot, but the best way to get good audio levels with little to no noise is to put your mics as close to the actors as possible.

As for your situation right now, after the fact, I think you're going to have to adjust each clip manually if you want consistent results. No software can look at waveform data and produce consistent levels throughout a project. At least none that I know of. Besides, there's more to getting proper levels than simply matching amplitude. There are a million variables. You just have to make manual adjustments and listen for what sounds best to you.

This is one big reason why it's best to get as good a recording as possible while shooting. The alternatives (extremely tedious tweaking on a microscopic scale, ADR) are time-consuming and may not even produce acceptable audio after all that work.

This is a lesson everyone learns the hard way, unfortunately.

[Jon Fairhurst]

Originally Posted by Jarrod Whaley

I hate to be Reverend Preachalot, but the best way to get good audio levels with little to no noise is to put your mics as close to the actors as possible.

Even so, you can't control the talent. They might drop a syllable. (A common one is to say "spore" rather than "sport".) They might also turn their head away from the mic.

In any case, if you use a clean track as your base, and mix a noise-reduced track to normalize the volume problems, you'll keep your noise floor from following your envelope.

Jarrod's right that you want to capture the best audio you can, but once it's captured, it's great to know how to patch it up.

BTW, as far as overall levels go, try to keep the peaks of normal dialog at about -12 dB. I've studied some nice DVDs, and this seems about right. Mix the music to taste. Feel free to let your peaks get within 1 or 2 dB of full scale. You'll find that you'll turn up your volume a bit to hear the dialog - just like on a good DVD. And when you need that explosion to knock people over, you'll have enough headroom to make a go of it.

BTW, you also asked about the normalize function. I always run my audio at 24-bits and normalize. It's just a single multiply, so any degredation is inaudible. The advantage is that you can see the waveform clearly on your track. I usually normalize to 1 to 3 dB below full scale. Don't go to full scale or the system might not handle HF overshoots properly at the peaks. If you can already see your waveform clearly, there's no need to normalize. Instead, work your gains and the volume envelope.

[A. J. deLange]

Originally Posted by Jon Fairhurst

BTW, as far as overall levels go, try to keep the peaks of normal dialog at about -12 dB. I've studied some nice DVDs, and this seems about right.

And it is about right theoretically as well. -12 dB FS is about the level at which an A/D converter will begin to clip a complex (as opposed to a pure tone) signal. The actual number depends somewhat on the number of bits in the conversion. A pure tone begins to clip at -3 dB FS.

[Bob Grant]

Originally Posted by A. J. deLange

And it is about right theoretically as well. -12 dB FS is about the level at which an A/D converter will begin to clip a complex (as opposed to a pure tone) signal. The actual number depends somewhat on the number of bits in the conversion. A pure tone begins to clip at -3 dB FS.

This might be true with VU metering but they're pretty rare these days. I can certainly run anything to 0dBFS without any sign of clipping, although it's a good idea to leave a little headroom (0.5dB) unless you've got some fancy tools to avoid intersample clipping in some D->A converters.
Of course one does need to be careful and know just how your meters are reading, if they're true peak reading meters then you're pretty safe with what I'm saying but the meters on some cameras I'm not too certain about.

[A. J. deLange]

Yes, good point. -12 dBFS RMS is the maximum load for an A/D. It is the load which allows full scale to be reached on extreme peaks but very rarely. An inertialess meter with instantaneous response will show an occasional FS flash FS with a -12 dB load.

[Jon Fairhurst]

I ran the dialog of our 48-hour film project entry with nominal speech peaking at -12 dB and it sounded great in the theater. There was room to get louder with forceful dialog, a fist hitting a desk, etc.

The film that screened right after us ran the dialog right up to full scale. It didn't have obvious clipping, but the audio came across like something you would hear during Americas Funniest Videos. The dialog was raw and harsh, and the ambient noise was very high.