Producing Pro Podcasts

Dec 1, 2007 12:00 PM, By Mike Levine

Hum Bug

If any of your interviews or VOs end up with significant background noise, consider using a noise-reduction plug-in. One affordable solution is BIAS Soundsoap 2 (Mac/Win), which lets you remove various noise types using automatic or manual settings. There are a number of noise-reduction software products on the market, including plug-ins like Soundsoap and Waves X-Noise and the standalone RX from iZotope. Some host programs, like Apple Logic Pro 8 and Sony Creative Software Sound Forge, come with noise-reduction plug-ins built in. These products do an excellent job of detecting and removing noise, but use them judiciously — if turned up too high, software-based noise reduction can make a voice sound very unnatural.

If you don't have noise-reduction software, there are often situations where you can remove problem noise using EQ. If, for instance, the offensive noise is steady at a fixed frequency — say, a 60-cycle hum — you can use an EQ plug-in to notch it out. Set your EQ for as narrow a bandwidth as possible (use the notch filter setting if there is one), set it for 60 Hz, and cut it to the maximum amount allowed. If you're not sure what the frequency of the offensive noise is, set your notch and sweep through the frequencies until you hear the noise lessen. You may not be able to remove all of it this way, but you can probably lessen it.

Another way to go, if the noise is of the low-frequency variety, is to filter out frequencies below about 100 Hz using an EQ plug-in. Your settings will depend on the frequency of the noise and the effect of the filter on your program material. You may lose a bit of bottom end from the voices in your Podcast, but if it gets rid of annoying noise, it's well worth it. Just don't overdo it and make the voices sound tinny.

Sweating the Details

When you listen back to interviews, you'll likely notice a lot of glitches including loud plosives, clicks, or just inarticulate phrasing (such as “um's,” “uh's,” and excessive “you know's”) and overly loud breaths. Unless you're going for an audio vérité, warts-and-all production, you'll want to edit out most, if not all, such anomalies.

I'll tackle breaths first. There are a number of ways to deal with them, including deleting them completely, or reducing their volume with automation or the Change Gain command (or equivalent). The key is to do it so that the result sounds natural. After a while, you'll be able to quickly spot breaths in the waveform display just by looking at them (see Fig. 2).

One effective method for getting rid of breaths in a natural-sounding way is to copy a little bit of room tone (ambient noise during a pause when nobody is speaking), and then paste it in to replace the breath. That way, the breath is gone, but there's still a pause between the words. If you simply cut the breath out and delete the space where it was, the speaker's phrasing may seem rushed and it may sound like you made an edit, which you want to avoid whenever possible.

If you completely reduce the volume of a breath, you'll hear an unnatural dropout at that point because there won't be any ambient noise. That is why using room tone is so useful (see Web Clip 1). MOTU Digital Performer has a feature called Smooth Audio Edits that places room tone (which you can designate or it can find for you) in between edit points automatically, with crossfades added.

If you want breath removal done automatically, Waves makes a plug-in called DeBreath, which is part of the company's Vocal Bundle. DeBreath is designed to automatically detect breaths and allow you to separate them from the program material.

Pop Goes the P

The noises made by loud plosives are distracting and should be reduced. Typically, you can't just cut them out, because the word will not sound right with a consonant sound removed completely. Reducing them is the best strategy. You'll find that it's pretty easy to spot plosives due to their distinctive, dense waveshape (see Fig. 3).

Here are the steps for reducing plosives:

1. Find the offensive plosive and zoom in on it.

2. Select the plosive part only (check by auditioning your selection).

3. Reduce the level of the plosive using either volume automation or the Change Gain command in your recording or editing software. Either way, reducing the level of the plosive by about 6 to 8 dB will usually do the trick, though sometimes you have to reduce it quite a bit more.

No matter what you're editing, always apply short crossfades at the edit point. Often crossfades will smooth out your edits. But if you have a click or pop at your edit point and crossfading doesn't help, you can often clean it up by changing your edit point slightly so that it happens on a zero-crossing, which is the point where the signal crosses from negative into positive or vice versa. Here's what you do: zoom in on the waveform until you find the zero-crossings for both sides of the edit. Then make your cuts so that the two newly joined sections meet at a zero-crossing (see Fig. 4). That will usually take care of the pop.

Level Playing Field

Another area that you're likely to have to pay a lot of attention to is levels. Human beings don't talk at perfectly even volume over the course of a conversation, so spoken-word recordings are going to have level variations you need to address. Whether it's voices trailing off at the end of sentences, loud interjections, or other inconsistencies related to volume, you're sure to find plenty of points during the course of an interview or voice-over where you need to make adjustments. You don't want the listeners to have to adjust their volume controls while listening to your Podcast.

A digital audio sequencer's volume automation is a great way to even out levels. You can also use compression for controlling dynamics, but be careful: too much compression can bring up the level of noise on spoken-word tracks. I generally lightly compress my voice-over and interview tracks using a plug-in that doesn't add a lot of color, such as Waves' Renaissance Compressor.