Got into a discussion with a teacher today about the proper level for recording audio, specifically dialogue. She was telling us that keeping it between -6 and -12 is what she does but who the hell speaks with that kind of dynamic range? I was under the impression that it was -12 to -24. Recording at the Dialnorm seems like it could decrease some noisefloor. Anybody have some recording level norms? I've googled it and come across a few answers but I want to see what this forum has to say!
The numbers you give seem to me to be when (post-prod) mixing. That's where I aim voice levels to be in a soundtrack. Talking purely about recording: I record dialogue as high as possible, obviously with safety. I aim to get the best wanted signal (voice) to noise (background) ratio. So the loudest moments will be just below 0. Depending on performance, voice can have a big dynamic range.
If recording too low, you won't be getting rid of noise, as you'll have to bring it up later in post. Not only will you be bringing up the background, but also any noise from the system. Rather record hotter, and bring it down later than the other way. It's kinda of a fine line though, cause you really don't want to be gaining up your recorder/mixer to the max either, as that will add way to much noise.
I'm just talking from my personal experience here, there are veterans hanging around here with tons of experience. Im curious as to what they say.
There are several factors involved; as with any type of recording there is 'no size fits all'. It depends on the subject being discussed, type of person and their animation whilst being interviewed, background noise etc. It's a matter of balancing maximum gain (digital amps are less noisier at max gain) without clipping, and whether you've opted to include the safety net of a limiter. If it's a one shot, with no chance of a pre-test then I suggest recording in the -12dB region and switch on the limiter, as once clipped the recording is ruined and opportunity lost. I cannot stress the importance of using max gain where possible for digital amps; they perform best at that point. It is often better using a pad and hi-pass filter to reduce the signal input in order that max gain can be used on the recorder.
+1 on Andrew's answer. When recording, the objective should be to get as clean, clear and present a recording as possible. A good boom op will be able to position the mic as close as possible to the talent (without being in the shot) in order to achieve this. It's in post where the dialogue levels will be adjusted to fit the desired norms. By aiming to record at a certain level you risk recording too low and, as Andrew says, the noise floor will be more present when mixing.
Inversely, you should always avoid recording clipped dialogue so don't record too hot!
Peronally I dont think you should see it as a wake up call! On location its WAY more important to capture the dialog without clipping than anything that has anything to do with technical noise floor.
Getting as close as possible is a given and minimizing real world noise is as well. but dont record hot as it WILL get you (and us in post) into trouble. If you aim to peak normal dialog at -6dbfs you WILL clip a bunch of material. Digital clipping sounds awful and is hard to remove cleanly even with izotope.
Using all the bits and worrying about system noise in location dialog recording is NOT time well spent, good gear today have enough headroom. Spend the time on removing actual issues in the room (such as reverb, sole noises, fans etc) instead of trying to maximise recorded levels. The dialog editors will love you. They will hate you for clipping the dialog...
Once, keeping general level between -12 and -6 was a necessity. We had only 16 bits (65535 individual steps with a possible range of 98dB), which meant the noise-floor was just around the corner. It was doable though mostly because limiting on the inserts, and as the gains where greater this way than the losses, we went with it. Actually, dialogue at that times where anything but sweet anyway, so noone noticed.
Now, we have 24 bit and beyond, 16777215 individual steps with a possible dynamic range of 144dB, with a dynamic range raised 6dB with each new bit. That means, to reach the same results as as a DAT on a modern fieldrecorder you can theoretically lower the gain approx 46dB. It's really not a good idea to do so as it would assume the recording quality (not to be confused with final quality) to be good. And it's not. We use the extra headroom and fidelity to make the movies sound better than they ever has and progress as an artform. I know many try to compare resolutions based on numbers, and that works more or less well for analogue formats (for example, cinema sound hasn't gotten louder with improved magnetic capabilities, they adopted a higher headroom instead), but in digital media the higher resolution makes a big difference, giving a more transparent, living and vivid result.
With that said, we now have the option to record even very dynamic sources with both full, undistorted, and relatively noise-free material by just adjusting to the new specs of modern day apparatus. Which sadly is something I've realized many old teachers find pretty difficult... Absolutely everything that affects the material degrades it. In a studio that's okay as you have control over it and can make more thought-through choices to reach transparent results if you can't undo them altogether, but on location, absolutely everything, be it an ugly brick-wall limiting, raising background noise/ugly mouth-sounds, or perhaps killing important transients altogether, will stay and can never be fixed to what it was supposed to be, alternatively take a tremendous amount of time could it be fixed at all. Besides that, on location there are no time to think things over too much, you must find a quick way to work and keep it on the safe side to avoid unexpected problems as far as possible. As very few people understands what we're doing, more responsibility to reach acceptable results lands on us, and that includes using our equipment as effectively as possible.
I follow pretty simple rules when I record dialogue nowadays; Keep centered level around -20dB, try to keep peaks below -3dB, preferably below -6 if possible, and keep any on-location processing, like filter and limiting, to an absolute minimum, if any. If I'm unsure how things will progress, I rather keep levels lower than higher to be on the safe side. I do of course always keep the limiter on anyway, as a slightly flattened peak still sounds pretty good even in editing if any peaks would cross 0, and they will at some time or another most definitely do, whereas a pure clipping sounds crap and might very well seriously F up a very important take, and make mixing less good. Something few directors are very forgiving about if there aren't a very good environmental or logistic reason...
In my experience after dubbing a lot of cartoons and stuff for tv that since tv in india is at -10 dB (sd cable or satellite), while dubbing in studio I set my preamp to get levels of max -10dB in protools for normal dialogues and upto -3 for any angry/shouting dialogue, cause everything is gonna be compressed/gain lowered later to fit with in -10 on tape
But for some major theatre release films, that I have started dubbing (with 2 mics) on with a reputed sound designer, all levels are hot, in the sense that I or the sound designer set the gain on the preamp to get levels as high as possible with out clipping.
In both cases-tv/films, for a loud shouting kind of dialogue, the artist is asked to step back a little, to get more voice projection without clipping. If he movies back too much he starts sounding hollow or I hear room tone.
In the studio recording voices for animation I'm aiming for around -10 or a bit lower. One important factor is knowing the actor and the script. If the script calls for sudden yelling or it's a very emotional moment, I am a bit lower. The same if I'm not familiar with an actor, or if I know it's an actor who's really all over the place. But when Seth MacFarlane steps in the booth, for example, knowing the characters and the script makes the levels very predictable, and I can safely aim for about -6 to -8 and get a particularly clean, hot recording.
I also aggressively ride the gain or fader as appropriate—in animation especially there is no set-it-and-forget-it.