Recording and mixing screams and whispers.

I

ironpony

Carbonite Member
I've been told before to increase or decrease the gain knob, if a person's voice get's too loud, or too quiet. Move the gain and not the fader. But I've noticed that by moving the gain down on a shout, that the shout actually records more quiet and it sounds fake, as oppose to the volume of the rest of the dialogue. Unless this is how it's suppose to be and I am just suppose to increase it in post. Should I decrease the fader instead, or still gain is the right way to go?

If gain is the right way, then what about whispers? But at normal dialogue, the gain is already at 75% and close to hissing. So increasing it, will only cause the whisper to have more hiss. So is fader the way to go for whisper, or still gain? Thanks.
 
What gain control knobs and what faders on what equipment in what specific recording scenario are you talking about here? Don't be so vague.

Assuming the "gain knob" you're talking about is one of the input channel trim controls on a mixer, of course turning it down is going to make a loud shout record at a lower level than it would without turning the gain down. That's what gain controls do. The channel faders are also gain controls for each channel. Think of the gain knob as a coarse level control, the fader as a fine level control. If you have to crank the gain knob up so high as to be on the verge of audible hiss in order to get proper levels of speech, you have the mic too &*%^ far away from the talent and/or it's not pointed directly at their mouth. The proper way to gain stage is first to get the mic into proper position, set the channel fader to the unity gain mark, adjust the gain control for a proper recording level on the mixers meters and leave it set there, then ride gain as necessary during the take using the faders. I would suggest just the opposite of what whoever gave you the advice you quoted told you - set and forget the gain knobs once for each setup and do all level adjustment fine-tuning during the take with the faders.

Part of the skills an actor needs is the ability to do a convincing shout without actually getting too loud and the ability to do a convincing whisper that is still audible. And part of the job of the boom operator is to know the script and tp position the mic so it's at the appropriate distance from the talent for the level of voice the talent is projecting, pulling back a bit for the shout so the mic doesn't overload.
 
Last edited:
Okay thanks. I am using the FR2-LE and I also have access to an H4n. In order to record normal speaking dialogue I have to set the trim (yes that's what I mean by gain), to 75% to 85%. Anything beyond that has hissing. I have to set the fader about two thirds in or a little less depending on the actors voice. So after setting the gain to 75% to 85% percent, I will just leave it there and ride the fader if I have been setting it correct so far. That set up is what's been working best for me, and giving me the cleanest audio, for normal speaking volume.
 
However it is managed (with separate channel gain and mix controls, or whatever), there are two separate (and sometimes seemingly conflicting) requirements here.

First: While recording the dialog during production, the objective is to get the cleanest recording possible. This means a signal with a good level (well up out of the "mud" of noise), while at the same time not so loud that you run into full-scale clipping. The proper settings for very soft dialog (whispering) and very loud dialog (shouting) will NOT be the same.

Second: While mixing the sound in post-production, the objective is to get all the parts of the dialog in PROPORTION to each other so that the drama is properly represented in the sound track, but within the same limitations of keeping the signal up out of the mud, but below the ceiling of clipping. Often some amount of limiting or compression or other processing is used to achieve this effect.

Note that one trick to achieving a wide dynamic range is to record the SAME signal (from your mixer) on two different tracks set at different record levels. One track will be set for higher record levels to properly capture the whispering without being buried in the mud. But this track will (predictably) clip during the shouting. So set the OTHER track for lower recording level to accommodate the shouting without clipping. But this track will lose the whispering down in the mud. During post-production sound mixing, you would chose whichever track gives you the clean recording and then adjust the levels in the final mix to suit the dialog.
 
The split recording is something we like in post. I don't do a lot of production work so I will bow to those who do, but I generally try to do as little "mixing" in production as possible. I will duck a bit if I know a big scream/ shout, etc. is coming up but the MixPre has really nice limiters and so I know I'm not going to clip out the recorder. But if you make adjustments during rehearsal and then between takes 9I find anyway) that I am not often making changes during the take.

I believe traditionally there was a lot more "mixing" in production. I'm not sure if new mixers are still doing a lot of level riding on the set. With analog it was an important tool to reduce hiss and noise but with digital recorders and clean preamps I'm not sure it's much of an issue today.

The reason I try to minimize moves is that I do a lot of post and one of the frustrating tasks is trying to undo bad mix moves from production. It's not just the level but the quality changes when the levels are moved around in production. Not huge if it was the right move but when it isn't, correcting it in posts can make the noise floor jump all over the place and then you are running noise reduction on bunches of clips to try to get them to fit with the rest of the dialog.

Since you are not tremendously experienced at sound mixing I would recommend you go easy on the moves. You want a strong signal with out any clipping. It should have dynamic range so don't ride the hell out of it. As a general rule over 75% on any gain control (trim / fader etc.) is going to add noise. Also as others have said a trim is a set it once kind of thing, it is designed to adjust the gain to match what the mic needs. The fader is what you are supposed to use actively. It will have (again a generalization but generally true) a much smoother and longer scale than a trim and you can make marks so you know where to hit.
 
Okay thanks. I know about the two channel safety thing, but I don't have the money to afford a splitter right now, and my FR2-LE does not have the ability to make a copy of the signal, on the other channel. I don't think the H4n does either.

I can't record whispers though without turning both the gain and the fader up all the way though. Otherwise the whispers will only record at 24db or less. I tried getting the actors to whisper louder, but that just came off sound unconvincing, acting wise. Same goes with recording very quiet things like very quiet sneaky footsteps or something. I have turn both all the way up but then have noise. And I hold the mic as close as I can. Accept for whispering where I still have to hold it a couple of feet away while shooting. Any thoughts on how to record those things without having turn both up too high?
 
ironpony said:
Okay thanks. I know about the two channel safety thing, but I don't have the money to afford a splitter right now, and my FR2-LE does not have the ability to make a copy of the signal, on the other channel. I don't think the H4n does either.

I can't record whispers though without turning both the gain and the fader up all the way though. Otherwise the whispers will only record at 24db or less. I tried getting the actors to whisper louder, but that just came off sound unconvincing, acting wise. Same goes with recording very quiet things like very quiet sneaky footsteps or something. I have turn both all the way up but then have noise. And I hold the mic as close as I can. Accept for whispering where I still have to hold it a couple of feet away while shooting. Any thoughts on how to record those things without having turn both up too high?
Click to expand...
Think about the psychology of the shot. The audience would only be able to hear a character's whisper in the real world if they were very close to him. This in turn translates to an ECU in the grammar of film. The tight framing of an ECU means that you can get the mic in closer, perhaps as close as 6 inches from the subject's mouth, without intruding into frame. If you're trying to frame it as a MS, the whisper would be inaudible to the audience if they were observing in real life. So come in closer with your shot and you should be able to get the mic close enough to pick up a realistic sounding whisper. Or have the character obscure his lips with his hand as he whispers in the other character's ear and record a wild line in studio for the whisper to add to the track in post. If your framing is such that you can't get the mic any closer than 2 feet, you're far enough back that the audience would be unlikely to hear the whisper if they were standing at the camera position anyway.
 
The manual is dreadful because it doesn't explain the mic trim function - but mentions it and the record level control together! The way you need to approach this is that the trim sets the pre-amp gain, and you want to use all it's available range. So for a mic that somebody is going to shout into, the gain will be set very low, for the whisper, it needs to be higher - the record level controls can then be working at perhaps 70-80% of full giving you some gentle adjustment up or down. Keep in mind that although perhaps not common for normal use, it could be necessary to fade out the recording - so if you had the input gains up full, with the hiss this produces, and the record level controls down very low, you'd have a very small range to fade. The idea is simply to use just enough pre-amp gain to give you a useful range with as little gain as possible. This also is important when faced with two inputs that need different gain settings - maybe a distant mic, and a narration mic? You really want to have the record level controls at about the same position for ch 1 and ch 2. On my recorder, it's a pain when you have to un-gang the concentric controls because you got the gain setting wrong. You'll soon get to know how far you can push the pre-amp gain before noise creeps in.
 
Okay thanks. It seems I have about 15-25% before the noise comes in. But even with the mic up close I still have to push over the noise line, just for the whisper to be recorded at 12db. And this is recording whispers close up. Unless I am setting it wrong?
 
Why should whispers reach -12dBFS? Normal speech should have peaks flickering at -12 or -10 dBFS but whispers, being quieter, would be lower.
 
Oh okay, that makes sense. Everyone way before said to make sure it records at least 12, but I guess they were not counting whispers.
 
j1clark@ucsd.edu said:
Where is your 'noise floor' relative to these -12 or -10 dBFS peaks?
Click to expand...
I don't have the OP's particular recorder so I have no idea what the noise floor might be. I'm going by the normal alignment practice of a 0VU tone at the input aligning to -18 or -20 dBFS. That puts 0VU speech flickering up around -12dBFS with some comfortable headroom for louder lines before you get up into the danger zone near clipping. While the OP is feeding his recorder directly instead of through a mixer the same principle applies ... normal speech hovering around -12dBFS on his peak reading meters gives the same comfort zone before clipping. I'm betting that the hiss noise he finds objectionable is an artifact of the poor headphone amp on the FR2LE and does not represent the noise floor of the recording itself. Even with his whispers are well below -12, I don't see that as a problem because they (should) stay soft in the final mix. He wouldn't bring them up to the same level as normal speech in the final mix because whispers are not, in fact, that loud. Not increasing their level would mean he's not increasing whatever noise is under them either.
 
Steve House said:
I don't have the OP's particular recorder so I have no idea what the noise floor might be. I'm going by the normal alignment practice of a 0VU tone at the input aligning to -18 or -20 dBFS. That puts 0VU speech flickering up around -12dBFS with some comfortable headroom for louder lines before you get up into the danger zone near clipping. While the OP is feeding his recorder directly instead of through a mixer the same principle applies ... normal speech hovering around -12dBFS on his peak reading meters gives the same comfort zone before clipping. I'm betting that the hiss noise he finds objectionable is an artifact of the poor headphone amp on the FR2LE and does not represent the noise floor of the recording itself. Even with his whispers are well below -12, I don't see that as a problem because they (should) stay soft in the final mix. He wouldn't bring them up to the same level as normal speech in the final mix because whispers are not, in fact, that loud. Not increasing their level would mean he's not increasing whatever noise is under them either.
Click to expand...

In 'researching' this question I've run across some 'recommendations' for voice over recording, which run like -70 dB for a 'quiet studio', and that parameter was listed in terms of defining 'noise floor' as 'electronics+ambient sound' noise, which actually is closer to my concept of 'noise floor' from other types of signal processing.

Another site recommended at 'quiet' to be at least -40 dB, and yet another stated 'quiet' as 20 db A SPL... hey I have an actual meter for that...

These sites also didn't make mention of the 'units'... so for the sake of forming my question... say they are all relative to 0VU... and so... that would be 'quite' is something like -88 dBFS or -58 dBFS...

Now with the above definition of 'noise floor' (I think one site had 'sound floor'...), a location production could never reach any of these 'quiet' levels. Maybe -40 or -30 dBFS... what this results in, a 'whisper' has to be some dB above -40 or -30 dBFS
and the question is 'what amount above', such that one can filter out certain bands of the ambient' to obtain a 'clean(er) whisper', but still the whisper has to be something greater than the ambient to induce in the human ear the sense that there is only the 'whisper' in the signal.

For example, relative to -40 dbFS does a 'whisper' of -30 or -25 'sound clean'... and if it is say -20 dBFS, that is 'close' to the -18 dBFS recommended for 'normal' voice.
 
You're comparing apples and oranges here. The indication of a recorder's meter as being "-xxdBFS" bears no fixed relation to a measurement of "xxdBSPL" for the sounds in the recording environment. The noise floor is the base electronic noise of the recording chain, not the sound in the environment where the recording is made. The numbers you cite for the sound pressure level of background environmental sounds have no relation at all to 0VU. 0VU can be thought of as the nominal operating level of an electronic circuit - a whisper can measure 0VU if the gain controls are turned way up while a jet engine can measure 0VU if the gain is turned way down yet the sound pressure levels of a whisper and a jet engine are wildly different from each other.

If the "noise" that is preventing the OP from recording the whispers clearly is environement sound levels, he's got other problems than the recording levels to worry about.

-20dBFS or -18dBFS is the recommended average level when recording of normal voice. The meters on a digital recorder or camera are usually measuring waveform peaks, not their average levels, and the peaks for voice waveforms are typically about 8dB higher than the RMS average level.

If I'm recording the line "Bob is a jerk" where the phrase "Bob is..." is spoken normally and the phase "...a jerk." is whispered, it would be very strange to ride gain to bring the level of the whispered phrase up to the level of the normally spoken phrase.
 
A good point of confusion there.

The OP is not experienced so maybe he is talking about environmental noise?

Clark, on a digital scale somewhere around -50dB to -60dB a track is essentially inaudible, if there is anything else going on. The threshold of hearing is something like -90dB on that same scale. But a few other things come into play. First no track should ever end up alone in a finished film, so you drop into the background sounds at a much higher level than what you can hear listening to a raw track through headphones.

OP, you should NOT be trying to record Foley (as in footsteps) on location. There is WAY TOO MUCH background sound to get useable results. They need to be recorded in a Foley stage (even if it's just a very quiet space you cobbled together) where you can get the kind of clean isolation that you will need to be able to add them to the mix.

Also if your doing a Hollywood fake where we get to hear the whisper from an unnatural distance then either use a wireless or ADR it. You probably know I am not a fan of ADR but this is one of those cases. Just make sure they turn so you don't have a clear shot of the lips and at that distance you should be fine. Then go into your "studio" (or quiet bedroom etc.) and record the whisper extremely close so you get that intimate "in your ear" sound and you should be good. If you have shot it right sync should not be an issue.
 
I'm really confused now. When RECORDING, there is no reason to not record whispers loud, peaking at the same levels as when you record the actors shouting - you are attempting to capture maximum dynamic range not reality! A recording using the full available dynamic range without ever hitting the full scale digital maximum is as good as it can be, quality wise. Common sense means that it's usually a bit less - but you are not attempting to balance out in the field, just capture. Back in the studio or in the edit, you balance the shout against the whisper - and dropping level is so much better than trying to raise it when the levels were poorly controlled when it was recording. Many people under record, level wise and spend time normalising in the edit - but sometimes this really brings up the noise floor!
 
paulears said:
I'm really confused now. When RECORDING, there is no reason to not record whispers loud, peaking at the same levels as when you record the actors shouting - you are attempting to capture maximum dynamic range not reality! A recording using the full available dynamic range without ever hitting the full scale digital maximum is as good as it can be, quality wise. Common sense means that it's usually a bit less - but you are not attempting to balance out in the field, just capture. Back in the studio or in the edit, you balance the shout against the whisper - and dropping level is so much better than trying to raise it when the levels were poorly controlled when it was recording. Many people under record, level wise and spend time normalising in the edit - but sometimes this really brings up the noise floor!
Click to expand...

This is my philosophy. Maybe it comes from my analog days where I pegged the meters on anything just to saturate the tape. It's better to turn something down than turn something up in post production. The problem with production, is if you drastically raise the level of a whisper, you are also bringing up the room tone, and that's a pain for dialogue editorial.
 
paulears said:
I'm really confused now. When RECORDING, there is no reason to not record whispers loud, peaking at the same levels as when you record the actors shouting - you are attempting to capture maximum dynamic range not reality! A recording using the full available dynamic range without ever hitting the full scale digital maximum is as good as it can be, quality wise. Common sense means that it's usually a bit less - but you are not attempting to balance out in the field, just capture. Back in the studio or in the edit, you balance the shout against the whisper - and dropping level is so much better than trying to raise it when the levels were poorly controlled when it was recording. Many people under record, level wise and spend time normalising in the edit - but sometimes this really brings up the noise floor!
Click to expand...

You are not recording a pop song, you don't need to be testing the limits of your bits. I tend to record hotter than most of the folks I've met who started in analog and I try to never get within 3dB of FS.

BUT again we need to define where we are coming from. I'm coming from long and short form film in the US. Your mention of "normalizing" makes me thing your coming from either music or the ENG/ video short turn-around world.

While I agree with Dave, I would always prefer to turn things down in post than try to boost them, what I really don't want is some production mixer riding the gain way up for the whisper and then back down for the dialog and then back up for the whisper. That sounds like AGC and is AWFUL. Now if the shot is all whispers sure pump up the gain, but if it's mixed hot and quiet then be gentle, or the post person will be sticking pins in a VooDoo doll with your name on it.

BTW your misusing the term maybe. If you ride the volume you are compressing the dynamic range not maximizing it. I think what you meant was you want to capture as much sound as you can. If your goal is to capture the full dynamic range you would leave the faders alone.

I agree to an extent. You want to capture as much usable dialog as possible with the best quality you can. That however is a dance. If you go crazy riding levels in production then a whole lot of time a processing is going to be needed in post and you have probably hurt your quality. On the flip side if the loud bits are clipped or the whisper is buried in hiss you have also hosed the quality in post.

Since there are some semantics and maybe production world differences going on we may be saying close to the same thing? Or we may have opposite views, not quite sure.

We can probably both agree though that an inexperienced sound person doing the field mixing is probably not the best person to try and "fix" the level on the fly.
 
You must log in or register to reply here.
Back
Top