Skip to content
Home » Dialog Editing: How to fill a scene with noise

Dialog Editing: How to fill a scene with noise

Why fill? Why noise?

Am I crazy? Why would I want to fill a scene with noise in the first place?! Great question. Noise is a part of life. Your brain filters is out in real time unless it passes a threshold you’re not used to, but importantly, your brain also expects it. If a scene sounds too clean, it can be distracting. Most importantly, if you have to remove a stray word or something (and you will), you’re gonna need something to fill that empty space, otherwise the audience will hear the noise disappear and it will take them out of the scene.

The way we dialog editors deal with this is simple. Your picture editor has probably done it before too, at least rudimentarily. We find a moment of recorded silence and we slot it into place. There are several names for this. You’re likely familiar with one or more of them. I prefer “fill”. It’s short, easy to type, and it’s unique. Here are some of the other names, and why I choose not to use them.

  • Presence – There are certain frequencies (around 3 kHz) that are often referred to as the “presence range.”
  • Room Tone – This can refer to the room tone recorded on set after a setup, or room tone recorded to be added in SFX track lay as an ambient bed
  • Ambience – Often refers to background sounds added in SFX track lay. Birds, oceans, starship engines
  • Background – Sometimes used interchangeably with Ambience. Other times, it refers specifically to background ADR/Walla. Human voices intended to be, well, background
  • Tone – Can refer to synthesized ambience tracks for alien worlds, horror, etc. Also refers to sine waves and pink noise used for calibration (collectively, “calibration tones”)
  • Fill – I’m not yet aware of a use other than for dialog editing. This is why I use it

How to fill (the traditional way)

We have a few tools to make fill. What does fill sound like though? What makes fill good? It should be clean (without much/any noise other than the broadband noise that defines the location). It should be long enough with nothing recognizable when repeated. Most importantly, it should match the sound of the shot you’re filling.

The best way to do this, in my experience, is to find chunks of silence near the moment you’re filling, stitch them together, and make it loop-able. Here are the steps I usually take:

Find silence that matches by looking in the following places (in order of priority):

Screenshot: ProTools dialog edit including ample use of fill between checkerboarded clips
Fill per shot crossfaded into dialog, used as bridge or elongated head and tail.
  1. Near the moment you’re filling.
  2. Within the same clip
  3. From the same location
    • Room tone recorded per setup
    • Other takes of the same shot
    • Other shots / room tone in the location
    • Synthesized fill (See next section)
  4. Sound library room tone (Absolute last resort, make it mono!)

Listen to the clips in sequence. You may need to trim heads and tails of each segment to make the fill sound consistent. Sometimes you’ll find that entire segments are unusable and need to be removed!

Note you may find a bunch of tiny chunks of silence between lines and right after “action!” Pull them all together with tiny equal-power crossfades. We often call this “stitching.”

Once it’s stitched, the fill might still be a bit imperfect. Some imperfections are tolerable. Others aren’t. Listen on loop and determine which sounds aren’t tolerable. Often, it’ll be fine and immediately usable. If it isn’t usable yet, I will bring the entire fill stitch into iZotope RX and use spectral repair to remove them. Bring it back into your DAW once it’s been cleaned.

If it doesn’t loop perfectly, you have two main choices.

  1. Duplicate the stitched fill, and reverse the duplicate. If it’s just broadband noise this will probably be sufficient. If not, you can do this:
  2. Make a cut in the middle of the clip. Now you have Part A and Part B. Move Part B to before Part A. Now the clip is loop-able. Add a final crossfade to the joint between B and A and the edit won’t pop.

Synthesized fill

Screenshot: Pro Tools and iZotope Ambience Match with fill
Dialog on the top, fill on the bottom. Naturually you won’t just lay the fill underneath, that’s for demonstration purposes only!

I try not to resort to this but it can work in a pinch. The first trick can be traced back to Shaun Farley’s blog post, a guest contribution by Douglas Murray, and even earlier. It’s been around since convolution reverbs have existed. Murray calls it “AIRFILL”.

AIRFILL: use a tiny sample of production room tone (I’m using this term to describe the uncut, unedited stuff, as opposed to fill, which is the final product) as an impulse in a convolution reverb. Feed the reverb some pink noise, and the output should be “perfectly” clean, steady fill. It’s great in a pinch, but is really sterile and can’t stand on it’s own. Good fill has a little… life, I guess? It’s hard to describe, but easy to hear. Check out the blog post anyway if you’re interested.

AMBIENCE MATCH: iZotope (praise be) created a module called Ambience Match. It’s quite expensive and I suspect it’s core is a simple convolution reverb, doing what AIRFILL already does, but it’s VERY convenient. Like AIRFILL, it’s “perfectly” clean and steady, and is missing the life that production room tone can give you, but it is still quite useful as a fill generator. Using the AudioSuite module in Pro Tools, you can learn the tone from a clip (even one with dialog, but best without), and then check “Output Ambience Only.” Usually you’ll have to reduce the volume by around 2-4dB, so they provide a handy trim fader, too. Add some handles so you can do crossfades, and you have yourself fill of any length you’d like, created offline. If iZotope reads this, PLEASE add some optional rudimentary EQ curves to the module. There’s often more low and high end than I want so I have to add an EQ afterwards!

I’ll usually choose Ambience Match as it’s easier to set up, but at least once, I’ve needed to use AIRFILL because ambience match just… it’s not always perfect! Still most of the time, the manual work above is best.

Wrapping up

Now you have a completely loop-able piece of fill. Rename it. If you’re making fill for 5B-T3.Boom, call it “5B-T3.Boom Fill A”. Add AF, IZ, or AM to the name if it was synthesized. If you need to make multiple versions for different moments (like when a plane flies by in the middle of a shot), “5B-T3.Boom Fill B” can help keep you organized.

Notice that in “Find Silence” I mentioned in 3.c you could potentially use tone from other shots in the location. This means your newly created “5B-T3.Boom Fill A” can now be used in the event 5A doesn’t have anything available in “Find Silence” 1 and 2. It shouldn’t happen often, but it’s not unheard of, so it can be a useful shortcut. Finally, as per 3.2, “5B-T3.Boom Fill A” will probably work well for a “5B-T2.Boom” clip as long as the background isn’t changing.