How Loud Should Voice Over Be?

[John Rofrano]

[Norm Kaiser] “I’m thinking I should set the audio levels on the preamp so the signal is strong enough to where the waveform thing on the audio track just stays under clipping. That is, the peaks should be as high as possible without hitting the ceiling and/or floor of the track.

Is this roughly correct? Or no?”

That’s correct. You always want your audio signals to be has high as possible without clipping. This will give you the best signal to noise ratio.

~jr

http://www.johnrofrano.com
http://www.vasst.com

[Graham Bernard]

[John Rofrano] “This will give you the best signal to noise ratio.”

Oh yes. It’s taken me the best part of 10 years to get this concept truly sunken in to my brain.

There is always Noise – it’s what is there so that the Signal can actually work. Manufacturers state S-N values and what is possible with XLR cabling and shielding and AC mains fluctuations and general PC noise. I know that’s simplistic in the extreme, but none-the-less it’s so that the clarity of the actual Signal:V/O, Music, Vox Humana etc, can be heard. Step One is to reduce the Noise to a minimum, and that I can really fussy and absorbed about. Step Two, get the best/affordable way to get the Signal into the PC. High S-N ratios equipment and cabling and shielding. So it’s the Ratio between S to N that’s most important. Do you have any noted values for this Ratio? 10:1? 15:1? 100:1?

Now, John, where does “Loudness” figure in this workflow? I’m assuming I get everything correct (is there anything that can assist/determine S-to-N?)and THEN pipe the whole stream via a Loudness meter/plug? This would be to comply with the rules and tenets of publication or broadcast.

Interesting post.

Grazie

Video Content Creator and Potter
PC 7 64-bit 16gb * Intel® Core™i7-2600k Quad Core 3.40GHz * 2GB NVIDIA GEFORCE GTX 560 Ti
Cameras: Canon XF300 + PowerShot SX50HS Bridge

[John Rofrano]

[Graham Bernard] “So it’s the Ratio between S to N that’s most important. Do you have any noted values for this Ratio? 10:1? 15:1? 100:1?”

The problem is that contrary to what you might think. signal-to-noise ratio is never expressed as a ratio in the specifications. For example the M-Audio Fast Track Pro has a S/N Ration of -97dB. The M-Audio ProFire 610 has a S/N Ration of -108dB implying that the noise floor is lower and therefore the signal to noise ratio is higher.

Here is the key concept: Since noise is a constant while recording you want the noise floor to be as low as possible and the signal to be has high as possible without clipping. In other words you can raise the signal without raising the noise while recording but once recorded the two are locked together during playback and raising the signal will also raise the noise.

Of course the “noise” we are more worried about in the field is the environmental noise around us (air conditioners, crowd noise, traffic) which must be dealt with by placing the microphone as close to the source as possible. In the case of a voice over, if you are in a vocal booth this is not a concern but if you are sitting in front of your computer as I am when recording tutorials, you want to watch the fan noise from your computer which is why I always build the quietest computer possible.

[Graham Bernard] “Now, John, where does “Loudness” figure in this workflow? I’m assuming I get everything correct (is there anything that can assist/determine S-to-N?)and THEN pipe the whole stream via a Loudness meter/plug? This would be to comply with the rules and tenets of publication or broadcast.”

Loudness is human perception. It’s often called “perceived” loudness. Loudness enters into the equation after you have recorded everything and are in the process of mixing it all together. For broadcast, loudness is measured over time. Broadcasters will specify an “average” loudness target. This makes it very difficult to get right because you must listen to a section, measure the average loudness, then adjust it, measure again, adjust it, measure again… until you get it right. After a while you get a feel for the program material you are working with and it gets easier but “loudness” is how the human ear perceives volume.

What I did when I was the finishing editor for Roger’s PBS show, is I added a 2:1 track compressor at -24dB on all the audio tracks. Then I set Wave Hammer up as a -10 dB brick wall limiter on the master audio bus followed by a Loudness Meter. Then I adjusted the volumes of the tracks to match the proper loudness target for PBS which is -24dB +/-2dB. If the audio on any track was uneven, I would increase the compression ratio to bring it back in line. Occasionally I would use a volume envelope to compensate but I tried to control things mostly with the compressors. This is the setup that Roger still uses today.

~jr

http://www.johnrofrano.com
http://www.vasst.com

[Graham Bernard]

Thanks John. Great reply. I understood all.

Grazie

Video Content Creator and Potter
PC 7 64-bit 16gb * Intel® Core™i7-2600k Quad Core 3.40GHz * 2GB NVIDIA GEFORCE GTX 560 Ti
Cameras: Canon XF300 + PowerShot SX50HS Bridge

[Mike Kujbida]

Grazie, if you have Sound Forge Pro 11, it now has a loudness metering option. I have no clue how to use it nor do I (thankfully!!) have to worry about it but it’s there if you need it.

[Roger Bansemer]

Thanks John. You’ve explained all this to me many times but it’s really good to get the explanation one more time.
Here’s one thing we do as a safeguard:
When recording in the field with a SLR lav mike, my camera will record on two tracks and unless I have a mic on two people I set it so I get a slightly higher or lower level on each track. That way if a track ends up being to hot I have the other track as a choice.
Of course if using two mics that’s not an option. If the interviewer and guest both have mics and one mic say for instance rubs against some clothing and causing some bad sound, I sometimes can use the first persons mic during that segment if they are standing close enough but have to be careful not to change the phase.
That happened on one of our shows and all sounded great on my PC but PBS when they reviewed it, heard nothing on that segment as the audio was out of phase.
I’ve know just barely enough now to get by with a passing grade when it comes to producing a finished program for PBS thanks mostly to Johns help.

Roger Bansemer – PaintingAndTravel.com

[Dave Haynie]

[Roger Bansemer] “Here’s one thing we do as a safeguard:
When recording in the field with a SLR lav mike, my camera will record on two tracks and unless I have a mic on two people I set it so I get a slightly higher or lower level on each track. That way if a track ends up being to hot I have the other track as a choice.”

I learned that as “run and gun” mode… with a single mic, you dial one to the level you think you want, the other down 12-18 dB, and you get an extra 2-3 bits worth of effective headroom on your signal.

Obviously, the best setup is to have a full-time sound guy actively monitoring the recording on 24-bit gear. Not always possible. But funny to watch me try to operate a boom pole and a camcorder (tripod-mounted, of course) at the same time 🙂

-Dave

[Dave Haynie]

And to the point of the question… you want the loudest signal that you can get that absolutely doesn’t clip (eg, it’s always below 0dBFS). I’m a big fan of not trying to get audio on a camera if I have another option… using my PC or my Zoom, I can put in a soft limiter, I have more resolution, etc. But basically, everyone agrees you want the best SNR you can get, which means the most signal possible.

Start with a quiet room. I did a thing last year with some voice-over, had the whole thing perfect, only then to notice that I had some annoying background noise in my PC room (mostly from the PC.. one reason the new one is crazy quiet). I was able to “fix in post” using a noise gate and a background track (duck that when you’re speaking, of course), and it sounded great, but I felt stupid for letting happen in the first place.

Depending on the situation, you can’t always be perfect. If you’re recording at 24-bit and producing for a 16-bit project, you have plenty of headroom… so in that case, try to get your peaks while you’re setting levels no more than say -12dBFS, and you’ll never risk blowing it out, even if you get a little excited while narrating.

For a voice-over, I’d probably compress 2:1 or 3:1, probably no more unless you’re a really dynamic speaker. I nearly always record “dry”, but this is one situation where it’s not likely a sin to use a compressor while recording, if you have one, particularly one with a soft limiter (my dbx 1066 is one of the few pieces of “classic — eg, real — audio gear I keep available, though with all the software tools and 24-bit recording interfaces, not so much these days).

If you’re compressing in a DAW, don’t just randomly choose a threshold, either… you really want to follow the signal with your compression. You can do this by ear of course, but I’ve gotten awfully spoiled by Izotope Alloy, which lets you dial this in graphically.

Then RMS normalize to about -16dBFS, give or take. Some goes as high as -10dBFS for voice… not sure I’m one of them. Whatever you do, being consistent is important.

-Dave