Lots of questions and not too many easy answers.
I'd start by normalising every clip, doesn't really do much as it isn't dynamic but at least everythings at around the same level.

You can try using eq to get rid of noises that you don't want but that's pretty limited, no magic bullet here. If you've got really loud noises between bits you want you can use envelopes to drop the level but if you don't have anything to fill it in it's noticable probably say it a noisy truck just gradually fade down as it becomes louder and then back up as it goes way. This isn't going to get rid of it but avoids another issue, when we hear something like a truck in the background there are other ues to the brain like doppler shift. We expect it to get louder, if it starts to fade away that draws more attnetion to it so you still want it getting louder but just enough, get the idea.

Sometimes you can patch in a bit of not so objectionbale noise in to cover the bits you've pulled out, when I shot film many years ago we always used to record some atmospherics onto the Nagra. We'd splice a loop of it on 16mm tape and run that through one track on the desk, very handy. Sorry getting nostalgic.

Basically though bad audio is much harder to fix than video, sound carries far more info to the brain than light, its an evolutionary thing, helped to stop us ending up as somethings dinner.


Anyway once you've got as much grot out of the audio as posible you can try the noise gate and compression. Be careful with the noise gate, I've heard even a lot of Hollywood stuff where I can really notice it, if your noise is above about -40 I wouldn't be using it unless you've got a lot of other stuff to mask the noise as well.

Compression really helps to firm up speech but again go for subtle, Sound Forge has some ever better tools, wish I could afford them!

Try to use eq and compression for the sound and level.
Regarding the background noise I suggest you to buy the noise reduction flugin. It works perfect for this.

if you want to get real crazy, look at a spectral analyzation of the files

and then figure out which freqs you want in them, (not real low or high ones)

and mold them accordingly

I haven't seen any way to do a spectrum analysis in vegas or acid. I don't have sound forge. I don't have the plug-in (is that the SF Noise Reduction Product, or is there a separate plug in?) I tried to EQ the noise out, but can't quite determine the right frequencies. What to do? (Besides spend $279 for NR)

It's ok to use different mics if you are using the different mics for a purpose of capturing different audio in different environments. Compression works the same way.... You can use lots of compression successfully in a quiet environment. But using compression in a noisy environment can cause a greater apparency of noise and a lot of background noise problems.

Normalization may or may not be the way to initiate the fix for your problems or bring all of your audio to an "equal sounding" level from clip to clip...

The reason I say that is because the dynamic picture and density of "wanted" audio will be very different and will have to be addressed based on those differences.

Normalization is an ok thing to do when you have a similar density between clips based on the dynamic or compressive result that has been recorded. But, if you have big differences in audio density, then you have to use normalization based on RMS instead of peak level. That can become very tricky.

Audio density is determined by more things than just how compressed or dense your "wanted" audio is. It is also determined by the frequency and size of your quiet areas where you happen to be finding unwanted background noise... and remember frequency actually means: "How often something happens"... We often forget that high frequency is referring to the modulated results of audio occuring more often...

The problem with using an RMS approach is; RMS is determined by an average, and silence (or the lack thereof) between words becomes part of that average....

Let's say a speaker's word; "HELLO" (if you only selected that word) had an RMS of -10db and that word occurs in a sentence; "HELLO, MY NAME IS CHARLEY". When selecting that sentence you found an RMS of -30db.

Wait a second! How did that happen?

Well...

#1. The sentence was spoken very slowly...

#2. with large pauses between the words...

#3. in a quiet environment.

If the same sentence was spoken by the same announcer in the same environment under the same recording conditions at the same volume, but the announcer spoke very quickly without pauses, you may end up with an RMS of -12db.

A very big difference indeed.

It would be fine to normalize to a peak level if you had used the same type of mic and mic pre audio chain. But, with different mics, you will have different RMS results based on the dynamic picture that the mic can reproduce. These dynamic variances combined with the acoustic dynamic picture, that was drawn by the environment you were recording in, can make a big difference in how loud or noisy one part may sound compared to another even if all of the peak levels are the same.

I'm telling you this so you will recognize/understand that you will have to treat your different audio problems concerning background noise and level with different approaches. There will not be a single group of fixes that you can apply to the different audio scenarios that you have found yourself faced with.

Hi,

Least hassle workaround:

-Normalize all audio,then cut out all the obvious noise that is not getting in the way of the important audio.-In other words: remove all the bumps,thumps,shouts and environmental noise you don't need.Ofcourse,by cutting there's always the chance that the result can sound pretty abrupt ,so only cut at places where you can get away with.
-Then use the noisegate to get rid of backgroundnoise which is not that apparent.

More hassle:
-those two steps mentioned above;
-remove all lows and highs you don't need for that track
-try to locate the frequency from the noise that gets in the way of the important data(speech and so on)and cut those frequency with the smallest bandwidth you can get away with.

A lot more hassle:
-all steps mentioned above
-rerecord (dub)the voices

Good luck,
Gerard