For my own future reference, but others may find it interesting: FilmSound is a site devoted to sound design and scoring for films, including foley and post-production. There's a large section on Star Wars that's fairly interesting, including the method of creating the lightsaber (a mic was placed inside a long tube, then that was waved between speakers playing the lightsaber "drone" to create its doppler-like movement).
Apart from producing the podcasts, recording voiceovers, and editing a couple of radio shows, most of my production work at the Bank involves supporting the video editors with their soundtracks. We use Final Cut Pro in the Multimedia Center, and although I'm sure it's a fine tool for video editing it doesn't seem to be very effective for more than the most basic audio work. For one thing, the effects need to be rendered before you can hear them, and I can never actually get any audible changes out of them (I'm probably doing something wrong, but the video people avoid the audio side like vampires in an Italian kitchen, so they're not much help). Soundtrack, the tool that accompanies Final Cut, is all well and good--but it's not really my cup of tea, and none of the editors want to put in the work to learn it.
So with all that in mind, here is a quick set of two tutorials for the common tasks that I perform while working my magic on soundtracks. I need to write a tutorial for my co-workers anyway, it might as well be now. Also, note that although Pro Tools can work with multi-track audio through the OMF/QT import part of the DV toolkit, you can also work with a stereo .wav of the whole soundtrack--and I often do.
1. Better Vocal Ducking:
When you bring in a voiceover on top of background noise, you want that background to get out of the way so that the vocals can be understood. That means lowering the volume, usually. Now, you can go through and manually adjust the volume using the track automation, but that's a pain and you might miss something. I prefer to do it through plugins. Now this is a pretty standard part of audio production, and you can google plenty of advice on it--assign a compressor to your background track, sidechain it to the vocals with a medium-high ratio, and its volume will automatically lower whenever the vocal track plays.
The problem with this, as my manager immediately pointed out when I first started using Pro Tools, is that it only ducks the volume right as the vocal comes out, and it sounds more natural if the background can start to fade just a little bit before the vocals actually enter, as if someone had anticipated the vocals. This also preserves the first word of the voiceover--it doesn't get lost in the slight pause before the compressor really kicks in.
I solve this problem, like most of my audio magic, with creative use of sends. Basically, you want to split your vocal track into two directions. Change its output to a bus, say Bus 1, and add a Send to a different bus, Bus 2 perhaps. Use Bus 1 as the background track compressor's sidechain input (the key icon in RTAS plugins will activate sidechaining if it's supported, and let you pick an input), but instead of setting the compressor to the usual settings, you want to give it a moderate attack and a very long release. I usually use the following settings (although I'm quoting from memory, so it may be off a little):
See what we've done? You've used the vocals to trigger a slow compressor, hopefully creating a fade, while simultaneously adding a delay to the audible vocal so that it'll arrive after the compressor turns down the background. You do need to be careful with timing, obviously, because the timeline display is now 333ms offset from your ears, but if you're after precision you should probably just invest in a plugin that supports look-ahead. This is the cheap way to do it.
2. Remove a noisy camera/audience:
Here's something to remember about audio: it's easy to add, but not so easy to subtract. Unlike video or images, you can't just cut a noise out, because it's part of the soundwave. Some tools, like Adobe Audition, will let you paint out chunks of the spectrum, but it's still not a perfect solution. And remember, sound is a representation of a physical phenomenon--in a lot of cases, it's the mic capsule or coil moving that is reproduced through your speakers. So physical movement or different sound frequencies can actually mask other sounds, because they're physically moving the mic and changing its interaction with the sound.
That's all very fun and technical, but what it amounts to in real life is that a lot of footage is shot in bad locations, through crappy equipment and sub-optimal mike technique. Maybe a speech was shot using the camera mike at the back of the room instead of plugging into the PA system. Maybe it was done on the move, and there's a lot of wind and crowd noise. Editors want that gone, or at least reduced, so you can hear the subject.
This is relatively easy to do. Remember that most of a voice's content takes place between the frequences of 100-5000Hz. Outside of that, you might miss some of the sibilant consonants, or low vocal rumble, but you'll be able to understand a person. Also, most electrical and camera noise takes place at the upper and lower limits of the spectrum. So to remove physical handling noise, like bumps, and boomy acoustics, I open up an EQ plugin and set a high-pass filter with a very sharp cutoff at around 180Hz, fine-tuning a little through headphones. I put a low-pass filter at around 6KHz, which minimizes clicks and a lot of tape whirr. If you add these through the AudioSuite menu of Pro Tools, using the preview function to listen before you apply them, you'll end up with a new, processed region that you can export back out for Final Cut.