If you need proof that editing is like turning a chunk of coal into a diamond, just listen to John Diliberto, the producer of Echoes. "That seven-minute feature that has three minutes of talking in it? That's edited down from an hour or so of interview material," he told me in "The Art of Podcasting."
Just as image editors like Photoshop gave us the power to zoom in on pictures and polish them to perfection, digital audio editors do the same for sound. In this article, I'll introduce some of the most useful audio editing commands and explain how you can use them to create your own sonic diamonds.
One of the most helpful skills you develop as you edit audio is the ability to "read" a waveform. With a bit of experience, you'll be able to look at a waveform and predict what kind of sound it represents and where the important changes happen. That mental map will enable you to quickly zoom in on the part of the recording you want to edit.
Figure 1, for example, shows the waveform for a famous three-word phrase, as displayed in BIAS Peak, one of the leading software audio editors. (See the sidebar for more.) The vertical axis, labeled 100 to 0 to –100, represents level, and the horizontal axis represents time. In other words, the height of the waveform at any given point corresponds to the volume of the sound at that instant. Notice how the first and third words ramp up to full volume gradually, whereas the middle word has a sudden, cliff-like beginning. It starts with a hard "G," whereas the other words start with a softer "Y" and "M." I like to imagine that the waveform diagram was produced by gluing a pencil across a speaker cone and rolling a long sheet of paper past it.
In our article "Killer Interviewing Tips for Podcasters, Part 2," Jack Herrington covered vocal editing in detail. But one area he didn't have space to address is plosives, the bursts of air that vocalists generate when forming letters like P, B, and D. Plosives smack against a microphone, producing an annoying thump. With a pop screen and good vocal technique, you can minimize their occurrence, but they still creep in. Here are four ways Jack identified to remove them from a recording. Because the first two work similarly, I'll demonstrate the last three.
Figure 2 shows a P-pop in a vocal file I received recently.
Figure 2. Notice the big spike where the audio level changes suddenly. It's an explosive P at the beginning of the word "piece."
Zooming in on the spike (see Figure 3), I deleted it.
Figure 3. We have now zoomed in and selected about 26 milliseconds of the P sound so we can snip it out.
Figure 4 shows the same area, tamed with a level change instead of deletion.
Figure 4. After processing with Peak's Change Gain command, the P-pop is now 12dB quieter.
Figure 5 shows the P after processing with a highpass EQ.
Figure 5. The height of the spike doesn't look much different after we filtered out the low frequencies, but it sounds quieter. Notice how it's denser horizontally. Low-frequency waveforms are wider.
The differences between these techniques can be subtle. For comparison, here is the sound again in its original state and then processed by EQ, gain change, and cropping, respectively.
One of the most important audio editing skills to master is volume fades, which help you transition between events without jarring your listeners. Here I'll cover fades in a musical context, using clips from Spencer Critchley's article "Country Music's Digital Surprise." Figures 6 and 7 show the clip before fading.
Figure 6. I find that musical fadeouts sound best when they begin and end on downbeats. Here I've inserted markers to identify the beginning of the bars in this clip. I'm going to create a fade between Markers 2 and 3.
Figure 7. Here's the bar of music before applying the fade envelope.
The simplest type of fade is a straight line (Figure 8), so I usually start with that.
Figure 8. Here's the effect of a linear fadeout.
When you need to make a quick stop, as in a radio edit, the slow fade (Figure 9) can work well.
Figure 9. Here's the effect of a slow fadeout, which is good for a sudden transition.
When you're fading to silence and want to give the impression that the song will continue forever, a fast fadeout (Figure 10) is usually best.
Figure 10. The fast fadeout has an exponential curve, which often sounds more natural.
Often, extending the fade into the downbeat of the next bar is effective. Listen to these examples:
When you have time, a long, fast fade sounds great:
Another trick I use is to run the fade command twice, producing a steeper rolloff. With the last example, that causes the sound to end in the middle of the bar:
To pull it all together, here are several fades on the final version of the song. (You may want to turn down your speakers a bit.) I also created a rapid fade-in:
Adobe Audition (Win)
Audacity (free; Mac/Win/Linux)
BIAS Peak (Mac)
Digidesign Pro Tools (Mac, Win)
Sony Sound Forge (Win)
Steinberg WaveLab (Win)
The best way to learn audio editing is just to dive in and experiment. Imagine what the recording engineers who had to slice magnetic tape with razor blades to make their edits would have given for an Undo button! (Not to mention a visual representation of the sound.) You'll find a list of popular audio editing programs in the sidebar. Most offer free demos, and one is entirely free, so there's no reason to suffer with lumps of audio coal when you could be making gems. For more audio editing tips, see "Killer Interviewing Tips for Podcasters, Part 2."
Return to the digitalmedia.oreilly.com