To overcome this issue, you need to understand and take advantage of one essential audio production technique: dynamic range compression. It does not need to be complex however: with a free software tool and no more additional work than a simple drag-and-drop, here is a practical tutorial to help you [...]]]> In order to achieve high-quality video commentaries that your audience will enjoy listening to, it is vital that you let the talent’s voice stand out loud and clear. But recording voice is more difficult than it may seem: sometimes it is too loud, and sometimes instead you can’t hear it well.

To overcome this issue, you need to understand and take advantage of one essential audio production technique: dynamic range compression. It does not need to be complex however: with a free software tool and no more additional work than a simple drag-and-drop, here is a practical tutorial to help you reach fantastic quality in your voice-overs.

Introducing The Levelator…

Any question or comment on this article? Come over discuss it on our Audio forum!

Why do we need to compress audio dynamics?

When adding a voice-over, your video typically ends up having 3 distinct audio components:

• The video soundtrack as was initially recorded by your camcorder (e.g. ambient sound),
• Background music that you may add,
• The commentary’s voice.

These 3 components fight each other for a share of the overall audio volume. For example, if at one point in time there is a loud noise in the ambient sound, it may cover the music or the commentary (which you don’t hear clearly anymore).

The issue you are facing here is two-fold:

1. In digital audio there is a maximum loudness threshold that you are not allowed to exceed. If your overall volume exceeds that limit, what then happens is called hard clipping: the audio sounds distorted, saturated. As a general rule, you must avoid hard clipping at all times.
2. The human voice has wide audio dynamics. It is very difficult to speak with a constant volume: sometimes you speak louder, sometimes softer, even within the same sentence (ordinarily people tend to end sentences with a softer tone). And some specific voice sounds – such as plosive consonants (‘p’, ‘t’) – introduce very rapid peaks in audio levels. So every time you pronounce those consonants for example, you are at risk of introducing audio saturation.

This is illustrated by the following diagram, which represents the 3 components in a typical voice-over soundtrack. Since voice is not constant (there are gaps between sentences), you may want to increase music volume between sentences, then decrease it again as your talent speaks. But even then, sometimes the voice is momentarily too strong and hard clipping occurs.

You could think that the obvious solution is to decrease the overall volume of voice? If you do this, you will effectively remove the audio clipping and distortion, but the soft parts of the recording may become even more difficult to hear (which is quite unpleasant for the audience watching your video). Because, again, the volume of voice is not constant and varies over time.

Benefits of voice dynamic range compression

The key here is to compress audio levels and decrease the dynamic range of your voice recording. In other words, think of it like reducing the variations of loudness by making audio levels more stable, more uniform over time. If you can compress voice loudness by making the softest parts louder, and the loudest parts softer, you will obtain 2 major benefits:

• The softest parts of the voice-over will be much clearer and pleasant to hear,
• You will reduce the risks of hard-clipping and saturation. When the volume of voice has been leveled this way, you have many more options at your disposal for mixing the voice-over with ambient sound and background music.

Enhance your voice recordings with The Levelator

Enters The Levelator. The Levelator is an adaptive audio dynamic range compression software, which improves the voice dynamics by performing automatic loudness adjustment. Although this very simple (and free!) utility was originally built for podcasts, it works equally well for voice-overs. You only need to drag and drop your raw voice recording (WAV file) onto The Levelator, and it produces another processed WAV file while leaving the original one untouched. There are no settings available, so you cannot tweak it to your liking, but this does not really matter: the default results of The Levelator are extremely satisfactory.

For example, here is a sample unprocessed voice recording. It is impossible to increase the overall volume because there are already a few audio peaks that are close to the hard-clipping threshold. And yet, there is considerable waste of dynamic range as the voice tends to fade towards the end of each word group.

Raw voice recording (click to listen to the audio)

Here is the same voice recording after it has been processed by The Levelator. Audio levels have been normalized: the voice sounds louder, and the end of word groups is much easier to hear. This makes a considerable difference when mixed with the ambient sound and background music, with splendid results and an excellent overall quality for the voice-over.

Voice recording compressed by The Levelator (click to listen to the audio)

Other audio peak limiter software

If you would like more control when compressing audio dynamics, you can also use Audacity which includes a compression filter. But it is more complex to use, with not-so-magical results.

But there is a much better solution, at no additional cost.

at no additional cost.

Any question or comment on this article? Come over discuss it on our Audio forum!

Microphones for PC sound cards have such a low signal, they require overamplification which brings signal-to-noise ratio (SNR) to about 35dB in voice recordings (YMMV). This results in a lot of digital noise contamination with a non-natural voice affected by digital artifacts. This could be due to the insufficient quality of motherboard chipsets naturally, so I replaced this with a full-featured sound card. Little improvement. This might be okay for Internet telephony (VoIP) such as Skype, but totally unacceptable for video productions.

I then found a very nice solution instead using my camcorder microphone for live voice recordings:

• You need a DV camcorder, or an HDV camcorder configured to record DV.
• Download and install River Past Audio Capture. The free trial allows for DV/Firewire recordings of up to 2 minutes in a row.
• Use your camcorder as is, or connect an external microphone — in my case, HDR-FX7E with RØDE Stereo VideoMic. Be sure to remove any tape from the camcorder (to prevent noise from the tape mechanism) and connect it to your PC through Firewire (IEEE1394).
• Use River Past Audio Capture to record voice-over WAV files, with live audio recording through the Firewire interface. Again, note that with an HDV camcorder this will only work after you have configured the menu to record DV (not HDV).
• Drop the WAV files onto your editing timeline.

From there, a couple of tips for best quality:

• You need to speak close to the microphone (I found 10 cm / 3 inches to be adequate).
• Standard wind protection is most recommended to protect against sounds from breathing at the microphone.
• Whether you set the camcorder for automatic or manual audio gain is up to you. AUTO guarantees similar audio levels for all recordings, at the expense of some loudness artifacts as you start speaking sometimes (because of the abrupt change in volume, therefore gain). MANUAL will avoid such inconvenience, but you need extra care and control in your voice to avoid audio clipping.

Overall, although professionals might call this a substandard cheapo hack, I found it extremely satisfactory with near-professional quality — more than enough for just any serious hobbyist.