Intro from Jay Allison: Transom is as excited as anyone about the possibilities of podcasting, especially the ability of artists and producers to reach their audiences without anyone’s permission. That freedom, however, does not obviate the need for quality—in content and in sound.
Yes, new technology has made a lot possible for regular folk, but the skills and ears of the audio engineer are not obsolete. If you’re venturing into this world, you need to know some stuff. Once again, Jeff Towne can help. In Part 3 of our series on Podcast Basics, Jeff covers loudness, EQ, noise reduction, and audio processing in general. It’s good to know this stuff, even if you don’t do it yourself, but just so you can talk intelligently to the tech folks you collaborate with. Good luck.
Get the Foundation Right
You’ve recorded a great conversation, or maybe you’re combining your voice with music, or sound effects, but it’s just not sounding “right.” That’s not unusual, there’s a lot involved in getting the right mix: choosing the right elements, balancing levels, fixing noises, and making sure the final production is at the right loudness.
The first rule of getting a good final result is to start with good raw elements. Try to get clean recordings to start with; don’t just assume that you can “fix it in the mix.” The first step is to pick the right microphone and to use good mic technique. (See the first article in this series.) Do some test recordings and listen back, then make some adjustments and try again, until your recording sounds good from the start. Does your voice sound distant and echoey? If so, get closer to the mic, or adjust the mic so that it’s closer to your mouth. If it’s still echoey, you might need to try some acoustical treatments. For ideas on that, read our article Voice Recording in the Home Studio.
Does your voice sound too bassy and boomy, or are you hearing distortion when you say words with a P or B or S sound? If so, back away from the mic a little, move the mic off at an angle. The most effective way to reduce P-Pops is to use a pop filter in front of the microphone. There are some techniques available to reduce P-Pops when mixing, but it’s much better to avoid having to use them in the first place.
The same goes for background noise: there is some amazing software that can reduce unwanted noises, but the results are never perfect, so it’s better to eliminate the noise before recording it. Turn off or move away from fans, refrigerators, ventilation, and other sources of noise. Similarly, if you have an interviewee or collaborator who is recording their own voice, ask them to observe the same rules.
Pick your other audio elements carefully: if you’re going to layer your voice over music, avoid vocals or saxophones or electric guitar leads that sound too similar to your spoken voice; they’ll conflict, and make it hard to understand what’s being said. The same thing holds for sound effects and other ambience.
Levels and Loudness
The most important part of achieving a polished mix is to keep your audio levels consistent. This is a difficult skill to master, but there are some technological aids that can make it easier. The first thing to keep in mind is that the most accurate measurement of the perceived volume is obtained by using the scale called “loudness.” There’s an in-depth explanation of the hows and whys of using Loudness here.
Many people have gotten in the habit of “normalizing” their audio in order to even out their levels, and hope that by maximizing the volume of each clip, they’ll all be the same loudness. But, if you’ve tried this, you’ll know that traditional normalization doesn’t really achieve this. Most normalizing processes look for audio peaks, the loudest moment in a clip, and adjust the sound file’s level to make that peak as loud as possible, without distorting. But it turns out that peaks do not correlate with perceived loudness very well. A stray cough, or clap, or click could force the level down more than it should go. Conversely, certain kinds of sounds that don’t have strong peaks can be made overly loud when using peak normalizing.
With the increased acceptance of Loudness as a standard, there are some programs that can do “Loudness Normalizing” which uses the more useful measurement of perceived loudness, and adjusts the clip’s level so that its average loudness matches an abstract standard. If you have a utility that can do this, it can be very useful.
The simplest way of using Loudness is to insert a loudness meter on the master track, and adjust the levels of each element up or down until the correct values are displayed on that meter. You can do this by adjusting the overall volume of a track, changing the level of a whole clip, or creating volume automation. There are some tips about mixing technique here. That column is about Pro Tools software in particular, but the basic concepts are pretty universal. Of course, you still need to use your ears to make the final decisions, but watching your loudness meter can help keep your mix consistent. One important note: the Loudness article referenced above describes radio broadcast standards, in particular the conventions adopted by the Public Radio Satellite system and other organizations supplying material for Public Radio.
Podcasting has different requirements. There’s no official standard for podcasting, but there’s a strong push toward using -16 LUfs as a loudness standard, rather than the lower -24 LUfs established for radio broadcast. Many podcasts are adopting the -16 LUfs standard, which will make the user experience better, as most podcasts will play back at the same volume.
Adjusting a mix to stay consistent at -16 LUfs is trickier than it seems. It might require not only careful adjustment of the level of each sound clip, but also some riding of the levels within the clip. On top of that, even with careful tweaking of each element, it might be difficult to achieve the -16 LUfs value without causing clipping, which will light up the red lights on your peak meters and usually causes an unpleasant crunchy distortion. The solution for this is to use compression and/or limiting, either on individual audio tracks, or on the master track, or both.
Compressors and Limiters can reduce the difference between the loudest and quietest parts of a sound file, making a mix more smooth and even. The short version is that using gentle compression on each track, set so the plug-in indicates about 4-6 dB of gain reduction on loud peaks, will make your mix sound more consistent. A limiter on the master track, also set conservatively, with a threshold of -3 dB or less, and an output level set to just below 0 dBfs (perhaps .5 dBfs or -1 dBfs) can usually control any stray peaks that could otherwise create distortion. There are articles with more details about compression and limiting here and here.
Most digital editing programs come with some kind of compression plug-in, although they vary in quality and ease of use. Hindenburg’s built-in compressor sounds surprisingly good and is very easy to use, with just one dial to adjust. The compression plug-ins that come with Pro Tools, Audition, Garage Band or other programs (as well as the many third-party plug-ins that can be added on to most programs) also give good results, although there are usually a few more dials and sliders to learn about. But the concepts are pretty universal: try a compression ratio of 3:1 or 4:1, then adjust the threshold control until you see the gain-reduction meter indicate 4-6 dB of reduction on the louder peaks.
A limiter is very useful on the master track, adjusted so that a Loudness meter on the master fader shows -16 LUfs. Some programs have built-in limiters, or there are many third-party limiters that people find useful. The Waves L-1 is one of the most used in the radio world, offering protection from audio clipping, while sounding clean and natural. If the Waves L1 isn’t in your budget, there are cheaper, or free, plug-ins that will also do a good job. LoudMax by Thomas Mundt is a good free limiter, and there are VST versions for most Windows programs, and AU versions for most Mac programs. (There is not an AAX or RTAS version for Pro Tools.)
If learning about compressors and limiters is too daunting, there are a few programs that can help you get to the proper levels. Hindenburg Journalist has built-in loudness normalizing, which is implemented in two different ways. The first is that the program will adjust the level of individual clips to match a loudness standard set in the program’s preferences, in fact it does that automatically (unless you turn it off) when you import sound into the workspace. You can also level a clip later by highlighting it, and manually invoking the level command. But most conveniently, Hindenburg recently added the option to export your mix with loudness normalization applied. So, if your mix is internally balanced, but you didn’t hit the target level, you can simply tell the program to adjust the loudness when you export it, and it will hit that target precisely. Loud and soft sections within the clip are not changed (with the possible exception of some peak limiting, if required), the level of the entire clip is raised or lowered consistently across its length. Beware: if your clip is not mixed well, with some sections too loud or soft, loudness normalizing will not fix that.
If you’re having a problem with peaks clipping, try setting the preferences to adjust your sound files to -24LUfs. You can mix at the radio broadcast standard of -24 LUfs, but then when you export, tell Hindenburg to apply loudness normalizing to the -16 LUfs standard and it will adjust the level, and apply peak-limiting if needed.
You can use that process in any program — mix to the radio broadcast standard, then use any of several programs to apply loudness-based normalization at -16 LUfs. There are several options mentioned in the Loudness article, including Adobe Audition, TC Electronic LCn and Auphonic.
Auphonic actually does more than just loudness normalization. They offer both a web-based service and desktop applications that adjust audio levels in complex ways. This processing can not only do loudness normalization, it can also make internal adjustments to even out internal inconsistencies in level. This is an automated process, and ultimately is no match for a human being making aesthetic decisions, but the results it is able to get, especially on dialog, are pretty amazing. You’ll get better results if you get your mix fairly close before using the processing , so that the program only has to make subtle adjustments, but Auphonic can effectively adjust the mix to be fairly consistent, even if you feed it something with widely varying volumes.
The service allows a lot of customization — of the kinds of processing it does, of metadata, and even of final product delivery to FTP servers, web hosts, and services such as YouTube, Amazon S3, Soundcloud, Blubrry, and others.
You can process two hours of audio each month for free with Auphonic‘s on-line interface. Additional minutes can be purchased as needed. Or you can buy a license for a desktop version: a personal license for non-commercial use is less than $100, a full commercial license is about $350.
Bring the Noise
Auphonic goes beyond volume adjustment, it can also do noise reduction. It does not offer many user controls for the noise reduction, it’s mostly all automatic, so it’s not going to be perfect for every kind of noise, but it can do a pretty remarkable job of reducing steady-state broadband noise.
Here’s some real-world audio from the field, recorded with a severe hum and buzz. This would be very distracting to leave untreated, and simple EQ won’t make much impact.
Here’s that clip, run through Auphonic’s noise reduction, set to “Automatic.”
There are several built-in and third-party noise reduction programs that can be used for more customizable noise reduction — e.g. if your audio suffers from an annoying hum, or rumble, or background hiss. Adobe Audition has sophisticated built-in noise reduction that offers a lot of flexibility in reducing broadband noise, such as ventilation whoosh or electrical buzz. The free program Audacity has a surprisingly decent noise reduction plug-in. You can “train” the plug-in to recognize a certain sound as noise by playing a short example of the noise alone. That noise profile is then used to eliminate that sound from the program material. There are only a few adjustments that can be made, but it can be effective. It’s hard to complain, because it’s free…
Speaking of things free to all comers...
Help Transom get new work and voices to public radio by donating now.
For Mac, there’s a program called Sound Soap. It’s been around for many years, and used to have an advanced “pro” version. In its most recent incarnation, it’s fairly stripped down and focused on ease of use. Its main target seems to be cleaning up audio for videos but it can be used equally well on audio-only content. Like most broadband de-noising programs, it has a “learn” mode, and a few knobs for manual adjustments. It has a separate hum-reduction module, which I didn’t find very effective, but the hum was reduced well by the other noise-reduction routines. There’s a de-clicker too, primarily designed to reduce surface noise from vinyl LPs. One unusual, and useful, feature is the “enhance” control, which adds some presence back to the processed sound, which can sometimes seem dull after the noise reduction is applied. Despite the relatively small number of controls, Sound Soap can do a surprisingly good job of cleaning up broadband noise. It’s relatively affordable, less than $100, and it’s pretty simple to use — there aren’t too many knobs and buttons to adjust, and what’s there is clearly labeled, with more details popping up if you hover your mouse over the control. Sound Soap does not work as a plug-in within another audio program, so you need to clean your audio before adding it to your project. It’s not as tweakable as some other programs, and is probably not going to be a perfect fit in all circumstances, but it can be helpful in many. Here’s SoundSoap’s de-noising, after using the “Learn” button, and a little additional knob twisting.
The noise-reduction software getting the most attention lately is Izotope Rx. It’s a powerful, and highly adjustable, collection of various noise-reduction modules. There’s broadband denoising (several types), but also sophisticated hum removal, de-clipping, EQ, and more. The “Spectral Repair” feature can reduce or remove very specific ranges of frequencies over limited time spans, while leaving the rest of the sound untouched, allowing for transparent repair of troublesome sounds like squeaks and beeps. There’s also an “Advanced” version of the software that offers additional controls, and a few extra modules, most importantly,”de-reverb” which can reduce echo and roominess in a recording.
The complex visual display is a real help in working on the audio, you can visualize problem areas, easily find areas of “pure” noise to use in training the software, and see as well as hear your results.
Of course, all this power comes with a cost, both financial, and intellectual. Rx4 (the current version at the time of this article) sells for $350, and Rx 4 Advanced for $1,200. And it’s not very easy to use right out of the box, although a bit of experimentation can often yield good results, even if you’re not exactly sure what all the sliders in the control panel do. But taking some time to learn the program can be very rewarding, it’s a very powerful tool.
Izotope Rx can operate as a stand-alone program, and in that mode, it’s capable of chaining multiple processes together, as well as batch-processing files, which can be a real time-saver if you have many separate sound files that need the same treatment. It can also work as a plug-in in most major audio editing programs. It’s available for both Windows and Mac, and offers plug-ins in AAX, RTAS, AU, VST and VST 3 formats. I’ve found that the plug-in version works fine for gentle processing, but for heavy processing, it’s better to use the stand-alone program, which offers non-real-time processing, which can often give cleaner results.
There are many other noise-reduction plug-ins and stand-alone programs available, each with pros and cons, but it’s worth remembering that in most cases, there is a sonic toll to the processing. It’s common to get audio artifacts when doing denoting, especially at extreme settings when removing severe noise, or when attempting very complex tasks, like de-reverb. The resulting sound can have a metallic ring, or gurgles and watery S sounds, like a bad Skype call, or a low-quality MP3. So, as stated above, it’s always better to eliminate the noise before you record it, rather than counting on fixing it with software. But if extraneous noise is unavoidable, it’s comforting to know that there’s some relief available through software.
The complexities of noise-reduction software can make your head spin. But there are easier paths to take. While it’s no match for the kinds of noise reduction you can get through the software mentioned above, using EQ – equalizing – can solve a lot of problems. The biggest problem solver is the High-Pass Filter: it can reduce bassy rumbles from traffic, wind or noise from machines such as refrigerators. It’s named High-Pass because it allows the high frequencies to pass through, while reducing low frequencies. This filter can also diminish P-Pops, and thumps from clumsy mic handling. It’s so useful a filter that the Hindenburg Journalist EQ inserts it by default any time you add their EQ plug-in (you can turn it off by clicking the little blue box on the left edge). If you have rumbles or hums or thumps (which not only sound bad, but can interfere with achieving consistent audio levels) try inserting a high-pass filter — almost any EQ plug-in will offer this option. Try sweeping the frequency up and down until you hear a reduction of problem noises. (You can do that in Hindenburg, and in many EQ plug-ins, by simply clicking on the curved line in the display and dragging left or right.) Be careful, if you make the frequency too high, it can make your sound overly thin and tinny, like the sound is coming over the phone.
You can also use a basic parametric EQ (the bell-shaped curve offered by many EQ plug-ins) to reduce other sonic problems. Try inserting the EQ, add some boost, then slowly sweep the frequency up and down, either by dragging the curved line in the display, or turning the frequency control in the plug-in, if there is one, until the offending noise gets worse. When you find the spot that sounds especially bad, then pull the line down, reducing the frequencies in that area of the sound. It’s often easier to get good results by cutting a problematic frequency range, rather than boosting other frequencies.
This takes some practice, and you need to listen carefully on good speakers or headphones in order to get the best results, but sometimes some gentle EQ can solve the problem better than more dramatic processing, such as the noise-reduction programs mentioned above.
The best rule is to be gentle: dramatic EQ adjustments often sound unnatural. Many EQ plug-ins have multiple “bands” allowing you to tweak more than one frequency range with one plug-in. Remember that you can also use more than one EQ in series to add more adjustments. Just make certain to be subtle with each EQ, or their effects might start fighting with each other.
Between evening out your levels, reducing noise, and adjusting EQ, you can create a polished sound that will sit well next to other podcasts. And remember, all these processors can be helpful, but the most powerful tools are your ears: use them as your main guide to what sounds best.
Output File Format
You should record your original tracks as WAV files, doing so will retain the best quality through the production process. But WAV files are too large to distribute as a podcast, so you’ll need to “compress” them into a more efficient file format: MP3 or AAC for delivery. The most common file format is MP3, and that’s still the most widely-supported by all devices that can play podcasts. AAC is becoming increasingly popular, and those files can be played on the vast majority of devices. AAC offers better audio quality at smaller file sizes, so it’s more efficient to store and to distribute, and it offers some additional metadata that MP3 does not, such as Chapter Markers, and embedded links and images. That said, those features are not widely used, and not supported on all playback software. AAC files are most often saved with the file extension .m4a. There are other audio file formats that could theoretically be used, such as Ogg Vorbis or WMA, but they are not supported by the players that most users choose, so those are not recommended.
Once you pick a file format, there are other choices to make about how the audio will be encoded. It comes down to a balance between file size and audio quality. MP3 and AAC are both “lossy” audio formats, which means that audio quality is reduced somewhat in the interest of making the files smaller and therefore more efficient to deliver over the internet, and more convenient for both you and your listeners to store. How much compression is enough? Too much? That depends a lot on the content of your podcast, and your own judgment about what sounds good enough.
Voice-only podcasts can usually be compressed more than music-heavy podcasts, with less noticeable degradation in quality. Mono files are half as big as stereo files, so think hard about whether your podcast needs to be in stereo. You probably want to keep your audio at 16-bit resolution, but some podcasters reduce their files to 8-bit to save space. Similarly, you probably want to keep your audio at 44.1 kHz sample rate, but some producers reduce the sample rate to 22 kHz to make the files smaller. Reducing either the bit-depth or the sample rate can have significant negative impacts on the sound quality, but it does make for much smaller files.
The biggest factor will be the “encoding rate” you choose: higher rates retain more of the audio information of the original WAV file, and sound better, but they’re bigger. Lower encoding rates will give you a smaller file size, and therefore you’ll need less space to store the files, and use less bandwidth to deliver them to subscribers. How important each of those factors is will vary depending on how long your podcasts are, what the content is, and how many subscribers you have. There’s no standard encoding rate used by all podcasters. Some use 160 or 192 kbps MP3, which is not a very severe compression, so audio quality remains fairly high. 128 kbps MP3 is more common for stereo content, 64 kbps MP3 is often used for mono spoken-word podcasts. AAC allows more drastic compression while retaining audio quality: a 64 kbps AAC-HE (high efficiency) stereo file sounds as good, or better, than a 128 kbps MP3, and is a smaller file size.
You may want to experiment with a few different kinds of file types and compression settings, and decide for yourself what format to use. A lot will depend on the content of your program and the relative importance of retaining fine detail. If your podcast uses music or subtle sound effects or ambience, you’ll probably want to lean toward higher encoding rates. If your podcast is voice only, and might not be starting with the highest-quality microphones, you might be able to go with a much lower encoding rate and still get acceptable sound quality.
Bandwidth, the amount of data used to deliver your audio to the end user, probably won’t be a big problem unless your podcast becomes very popular. If you have a few hundred, even a few thousand subscribers, the bandwidth used to send out that audio will likely be covered by the basic contract with your web host, or a specialized podcast hosting service. That said, be sure to check to see what your hosting service provides, and what happens if you do exceed the agreed-upon bandwidth. You could potentially be charged for extra data use if your podcast is downloaded many times. This becomes a big issue for the most popular podcasts: with hundreds of thousands, even millions of downloads of each episode, bandwidth charges can get quite high, and the motivation to keep file sizes small is great.
You could have worse problems: if you have so many subscribers that you need to worry about bandwidth, you’re doing something right, and you might be able to generate some revenue through ads, or underwriting, or other support.
The actual encoding of your files can usually be done directly from your editing program: most of them will have an option to export your final mix in a few different formats. Or, you can always export the mix of your show as a WAV file, and use a separate encoding program to convert it to MP3 or AAC. There are many stand-alone programs that can convert audio from WAV to MP3, some using custom algorithms and promising higher quality, but those tend to be expensive. The simplest way to convert a WAV to MP3 or AAC is to use iTunes. Import the WAV file into your music library. Make sure that the track information, the Title, Track, Album, and other metadata is displaying the way you want it. If not, add or change that information now, before encoding the file, so that the information gets embedded in the MP3 or AAC. Then go to the iTunes Preferences and find “Import Settings.” This gets moved around in various versions of iTunes, but it’s usually in General Preferences. Set that Import Preference to whatever values you’ve chosen, for instance, AAC, High Quality. You may need to choose “Custom” to find the exact settings you’ve chosen, or you can use one of the more popular presets they offer. Some versions of iTunes have a setting called “Spoken Podcast” which creates a very small file size, and even reduces the sample rate to 22 kHz. After you have that preference set correctly, go back to the music library, highlight the WAV file of your podcast, then go to File>>Create New Version. That will create a new copy of your file, converted to the smaller format, ready to upload. That file will be saved in your iTunes music library. If you’re not sure where that is, highlight the file in the iTunes library, then use File>>Show in Finder or File>> Get Info to locate it.
In the next column, we’ll discuss where to store your audio, and how to create the RSS file that will automatically distribute your podcast. But before you send any episodes out into the world, make sure they sound good!