I'm not going to cover proper sound capture techniques on set other than to mention the traditional bits. To wit: Microphone as close to subject as the shot will allow, you are capturing only the voice of the actor, not everything around them, wear headphones while recording and listen to what you're getting, watch your levels, keep them from peaking as digital will simply lose the sound that "clips". Capture any environmental sounds that you feel you need separately in 30+ second sound bites and sound fx from before the sound until after it's faded.
This particular tutorial deals with the bits that are ignored in sound; Post Production and Sweetening.
Understanding and Visualizing Sound:
Sound is made up of a range of differently pitched sounds (called frequencies) that work together to create the sounds we hear. Whether that be a train whistle, a waterfall, thunder or a person's voice, they are all comprised of waves of sound at different frequencies hitting the ear drum.
The human voice occupies a narrow band of these frequencies. Cutting out the frequencies not used in the vocal track will help to isolate your actor's voices and give you more control in post over the audio in your project.
To see your sound, use a spectrum graph rather than the waveform view of your soundtrack. The waveform so common only shows time vs. loudness (amplitude), whereas the spectrum graph shows time vs. frequency with amplitude shown as color within the graph. This 3-dimensional graph will allow you to see where the sounds in your recording occur. Most graphs show a linear scale for frequency, wee prefer to look at it as a logarithmic scale (in Soundtrack Pro, the waveform/spectrum buttons are in the upper right hand portion of the main window and switching to log is as simple as right clicking in the frequency scale and selecting logarithmic).
High Cut, Low Cut:
To eliminate as much background noise as possible from an audio track, apply a low cut at adjust it higher until it just starts to make your vocals sound hollow. Then apply a high cut and lower it until the same happens at the other end of the voices.
EQ Tricks, Notch Filter:
On set, we try to eliminate all sounds other than our actor's voices. We do this by using very sensitive microhones with a huge ability to reject off-axis noise. We try to get home and business owners to allow us to turn off any noise makers on set, refrigerators, ceiling fans, humidifiers, A/C...
These don't always work, either something gets missed or the owners won't allow refrigeration/ cooling to be turned off (like in an open and operating restaurant). In these cases, making sure the microphone is pointed in a way that minimizes these noises can be utilized (front and rear of the microphone perpendicular to the offending noise source and not toward surfaces that would reflect those sounds.
To find these frequencies, look at the spectral graph, horizontal lines relate to sound with a consistent frequency (like a humming refrigerator). To eliminate this humming, add a notch filter to the track. There are 3 controls: Frequency, Q and Amplitude. The frequency should match the sound you're going to remove, the Q adjusts the amount of frequencies around the target that are affected and the Amplitude is the volume adjustment.
Start by setting the Q as high as it goes, do the same to the amplitude. While previewing the sound (looping around the bit you want to fix), scrub the frequency until the sound becomes painfully really loud. Once you've identified the frequency this way, drop the amplitude as low as it goes and the sound should magically disappear. You may need to adjust the Q wider to cover the adjoining frequencies to fully remove the sound.
Lather, rinse repeat keeping in mind that we prefer not to negatively impact the subject of the recording. If two sounds are in the recording at the same frequency, the loudest sound will drown out the same frequency in other sounds in the recording. So if you do have to poke a hole in the subject of the recording, make sure to find ambiance, sound fx or music at those frequency ranges to plug the holes you create now.
This should get you started presenting cleaner sound in your productions without too much hassle. The frequency scrubbing technique will work evern without the spectral graph and allow you to do all of this sound work within your NLE if necessary (although it'll take more time to render, so account for that)... but if you're taking the time to learn all kinds of Special FX tricks, you may as well learn these tricks as well. It's all just operating on digital information anyway, and all of the time you take in post will make your results more and more professional.
This installation deals with removing sound. Later, we'll cover adding sounds to make your overall soundtrack richer and more professional sounding.
This bit is golden sweet sauce, practice is frequently! Watch out though, start by turning the volume DOWN cause you might hurt your ears.. seriously..
Start by setting the Q as high as it goes, do the same to the amplitude. While previewing the sound (looping around the bit you want to fix), scrub the frequency until the sound becomes painfully really loud. Once you've identified the frequency this way, drop the amplitude as low as it goes and the sound should magically disappear. You may need to adjust the Q wider to cover the adjoining frequencies to fully remove the sound
So last time, we talked about sound in your film as a single entity... Now we're going to concern ourselves with thinking larger. Isolating the actor's voices has hollowed out the background and even gone so far as to make some sounds go missing.
Let's look at what we've got. We've pulled in all our clips, we've edited them into a brilliant short film and we've filtered out all of those pesky background noise so we have pristine dialog. I'm going to have you go one step further now and actually cut out any audio that isn't dialog entirely... all of the bits between the actors talking needs to go away.
Each actor gets their own mono audio track. I would recommend exporting the whole film at this point and pulling the audio in an uncompressed format into a dedicated audio package. I'm using Soundtrack Pro, but the techniques will be nearly identical in any audio editing package out there as they all emulate old analog equipment techniques.
We'll be working with tools that may be new to you, I recommend you play with every knob and slider in each piece of the software you see to figure out what it does to the sound. In the matrix, we have undo - so there is no risk of completely horking your soundtrack just by twiddling knobs. Our editing should all be non-destructive!
Since our focus is on getting the highest possible production quality out of our soundtrack, I'll fall back on the same equation I've been using for years in all of the articles I've ever posted; Time = Money. Since I don't have money, I know I can spend time to get the same results that money would buy for me... I just have to be willing to work for it.
With the audio pulled in, separated (with the razor blade tool in most packages) into dialog and everything else, and even farther by cutting each actor's lines separately, which I will normally do by giving each actor their own clip in my video editor, then export each of their dialog separately as AIFF files which I then line up on separate tracks in Soundtrack Pro.
We now have the beginnings of a well rounded sound design. Since the actors are the only thing we are concerned with recording on set, the rest will now have to be added back in. This can be as artistic or realistic as the story calls for. I would even go so far as to say that other than dialog timing, everything else for your audio should be done in a program with stronger audio tools than your video editor (although most of what we cover will work just fine in there - but we want the highest quality results possible, and the tools are available).
So let's start by taking our freshly mown dialog and applying the high and low cuts from the first installment of this series, then EQing the background cruft out of it all. The high and low cuts can generally be applied track-wide separately from one another for each actor to make sure their voice is prominently separated form the background noises. Separate tracks allows us to be even more precise with the high and low cuts as different people's voices inhabit different frequency ranges.
Next, we'll balance the volumes of the individual actors to one another watching our levels to make sure they stay right around the -6db mark on the meter, never peaking above 0db - in fact safer yet would be to use the -6db as your peak level for the loudest bits. Here we can be a little creative as well and use the a limiter to make louder parts slightly less loud by "attenuating" them as their level increases. The compressor/limiter specifically applies more dampening to a sound clip as the volume increases, and we can control how much dampening gets applied and when. I won't go into specifics, but generally, you can select a highest level for the filter and it'll be set right... the change will be instantly noticeable in smoothing out the levels in your clip. I tend to apply one of these filters per dialog track after balancing the rest of the dialog to each other. This allows me to ignore screams and louder bits (high points in the waveform) and concentrate on the rest of the dialog as a whole.
Having pulled out everything but the voices, we now get to start piecing the sonic world back together. I like to start with easy to recognize things. Footsteps specifically. Many sound libraries have footsteps in them, the problem is that the foot falls don't always match the speed of your actor's gait. This is easily corrected by understanding our audio in a visual manner. If you look at the waveform, you'll notice thicker bits and thinner bits. This is a graph, left to right is time and up and down is volume. For things like footsteps, you'll tend to see regularly spaced spikes in the waveform that coincide with the foot contacting the ground.
My general work flow for these is to watch the video window in my audio package and use the arrow keys to nudge forward one frame at a time until I see the weight of my actor shift onto their foot. At this frame, I'll place the first wave spike in the footsteps audio clip. I'll then advance the footage a frame at a time until I see the weight shift onto the other foot. The razor blade cuts just before the next wave spike in the audio clip and I slide the footstep back to line it up. I don't worry about it overlapping the other footstep I just placed as the loudest sound at a specific frequency will overwhelm and hide the lower sounds at that frequency. Since the footsteps are all going to be roughly the same frequency, the audience will never hear that the old one got cut off as their ears will be hit at that same moment with a new sound... yay physics!
Lather, rinse, repeat! Then do the same thing for each actor with a new track for each of them, then same for the hands. Each time they touch something, there should be some kind of a sound, try to find something appropriate for the real world interaction you're showing on screen. Then sound effects to match the action outside the characters, cars passing, dogs being seen barking, children playing in the background. Each of these things builds a new layer into the world of the film.
Next bit we'll cover will be the environmental sounds and how to deal with room tone if we hear holes in our sound track.