Noise and Mixing overview
We are now deep into the audio post production process. The overview in part 1 laid out the case for taking audio seriously in the first place, the preproduction checklist in part 2 set us up for success, the location sound primer in part 3 spelled out what to listen for on the set, and the dialogue transfer and edit process explained in part 4 prepared us for the next step in the audio post process. From here we move on to the last big parts of the dialogue process - noise reduction and mixing
If you plan on doing the noise reduction yourself I would again recommend the book Dialogue Editing for Motion Pictures: A Guide To The Invisible Art by John Purcell. The latter half of this book goes into great detail about the art of noise reduction and dialogue mixing.
All dialogue tracks are going to have undesirable noise to some degree. Even projects shot a on soundstage can come back with various blemishes that will need to be removed before the mix. The process of reducing or eliminating the noise in a track is a delicate dance between what you have to do (because that is the take that you need to have in there), what you can do given the tools available and the type of noise being addressed, and what you can get away with as far as noise being left in, or artifacts of noise reduction.
Its very important to listen on a high resolution system when doing noise reduction. Small computer speakers or noisy rooms make for terrible noise reduction decisions that you will later regret, so make the effort to find a respectable playback system and a very quiet room when making noise reduction decisions. Also remember to listen loud enough to hear the artifacts you may or may not be creating.
We discussed in part 3 what kinds of noise you will and will not be able to get rid of during this part of the process. Of all of the tools that you have available for the purposes of de-noising production audio, your EQ is going to be the most useful. EQs can lop off all of the low end in a recording that doesn’t actually represent anyone’s voice, and can be used to surgically get rid of certain static sounds like electrical buzzes and light hums.
The second best tool is a multiband expander like the Waves C4. Multiband expanders break the audio into multiple frequency spectrums and then can turn down the parts that aren’t actively part of the dialogue being spoken by a few decibels. They aren’t panaceas, but they are often part of a good dialogue cleanup processing chain.
Noise Reduction Software
The last and most dangerous is tool is the actual noise reduction software. Noise reduction software works by sampling the offending noise and then using that sample to remove all of the audio that resembles the noise footprint. Noise reduction software can be very useful when used in moderation, but it won’t solve big sound problems and can often leave digital sounding artifacts behind as the noise in the recording shifts and moves around.
When using or evaluating noise reduction software keep in mind that your audiences will be far more forgiving of some extra noise being left in than of some unnatural sounding digital artifacts being introduced. The reason is because the noise that the world makes around us generally sounds natural and can be easily tuned out by the brain and forgiven. By contrast, noise reduction software artifacts are very unnatural sounding and can draw attention to themselves even in small doses.
The general rule of thumb with regards to noise reduction is that less is more. If you find yourself or your editor doing wild and weird sounding EQs, if you hear a gate or expander chopping in and out, or if you hear digital noise reduction artifacts, then you should take that as a sign to back off on the processing a little.
Even if you aren’t doing the dialogue cleanup yourself, its important to be able to make good decisions as to what is and is not acceptable in your film. Too much noise left in and you risk your audience missing or mishearing a critical piece of dialogue. Too much reduction and you risk drawing attention to the dialogue edit and breaking the illusion.
The last step is dialogue mixing. Most digital workstations don’t come with built in VU meters, instead providing digital peak meters. I would highly recommend either working with outboard VU meters or finding the VST plug in equivalent. The reason is that digital peak meters don’t give you any information as to how loud something actually sounds, and really only provide insight into whether something is about to digitally distort or not. VU meters measure average loudness and are a much better representation of how loud something seems to be.
For distribution methods that require a tighter dynamic range such as internet broadcasts and television spots you should generally set your speakers to a pretty quiet level and leave them there while you are setting levels. This will cause the softest parts of the dialogue to fall into the noise floor of your room and force you to turn them up a little more to compensate. For distribution methods that can allow more dynamic range such as DVD releases the speakers should be at a relatively louder level so that the nuances of the speech can be better preserved. The important step is to find the speaker level you’re mixing at and lock it down. This will force you to use your tools to move things to where they sound comfortable instead of your speaker levels.
If you have an SPL meter (the Decibel app for the iPhone 3g is great and only $0.99) then you can calibrate your listening environment by playing pink noise through each speaker, one at a time, and setting the level on each to be exactly the same loudness. 74 dB SPL is a good reference level for television and internet mixes, and 79 or 80 dB spl is good for DVD mixes.
Now that you have an actual useful measuring device the general idea is to set all of the dialogue at or near zero VU, while leaving your speakers set at a level that doesn’t change. Different applications will call for different levels of dynamic range, but generally speaking the average level of someone speaking should be right at 0 VU on your meters. Often compression is used to keep the whole edit within a tighter dynamic range, but be advised that the time spent volume-mapping instead of just clamping a compressor on will pay huge dividends in the long run. Over-compressed dialogue can sound crunchy and can become fatiguing quickly. On the other end of the spectrum, unmanaged dialogue becomes impossible to mix because the softest parts are just too much softer than the loudest parts, meaning that you don’t have a reasonable place to put the music because it will drown out words here and there.
Focus on the Dialogue
When mixing always remember that dialogue is the most important part of the mix, and consider the environment that your audience will be in when watching your digital video production. Mixes for laptop speakers will work differently than mixes for theatrical release, even if all of the content is the same.
The dialogue editing and mixing process is a delicate one. Even if you are not doing the dialogue edit yourself you should be aware of the various steps of the process. This will enable you to deal with problems proactively as well as realistically. There are some issues that are easy to solve during this process and others that are just not going to get fixed no matter what you do. Know the difference so that you can make good decisions in the picture edit, the ADR schedule, the overall budget, and the time allotted.
This post is part of the series: How good audio can help your digital video production
An in-depth step by step guide to audio for your digital video production. Learn about audio pre-production, location sound and post production. Learn what to listen for and what questions to ask, even if you’re not the audio guy on the set.