A party upstairs. A radio in a parked car. A band heard from the street. Here is what is actually happening acoustically, the settings that get you most of the way there, and the four things that make the difference between convincing and obviously faked.
Written for people cutting picture. Works in Premiere Pro, Final Cut Pro, DaVinci Resolve, Pro Tools or anything else with an EQ and a reverb.
A wall is a low-pass filter. That is the whole thing in one sentence, and almost everything else follows from it.
How much sound a partition blocks depends on its mass and on the frequency. The relationship is close to 6 dB more isolation for every doubling of frequency. A typical insulated stud wall might block around 20 to 25 dB at 125 Hz, around 40 dB at 1 kHz, and rather more above that. So the bass gets through and the top does not. That is why the neighbours' party is a thud rather than a song.
Two things fall out of that, and both of them are where people go wrong.
The slope is gentle. Six dB per octave is a first-order filter. Most people reach for a steep 24 dB per octave low-pass, which behaves like no wall that has ever been built. A gentler slope over a wider range sounds far more like a real room.
The bass is the effect. The instinct is to make it small and thin, so people high-pass the life out of it. Do the opposite. The low end surviving while the top disappears is the entire signature of sound through a wall. Cut the lows and you get a telephone, not a party upstairs.
These are opening positions, not presets to trust blindly. Set the low-pass first, get the level right, then add the room. Every figure assumes a gentle slope, 6 to 12 dB per octave.
| The shot | Low-pass | Level | Then |
|---|---|---|---|
| Next room, door closed | 500 to 800 Hz | 25 to 30 dB down | Small reverb of the room you are standing in |
| Next room, door open | 2 to 4 kHz | 12 to 18 dB down | Mostly room, very little filtering. You hear the space, not a wall |
| Solid wall, no opening | 250 to 400 Hz | 35 to 45 dB down | Almost no reverb. Nothing bright enough survives to reflect |
| Party on the floor above | 300 to 500 Hz | 30 to 40 dB down | Structure carries the low end further than air does. Keep more bass than feels right |
| Down a corridor | 3 to 5 kHz | 10 to 15 dB down | Long, narrow reverb. Filtering is minor, the corridor does the work |
| From the street, through a window | 600 Hz to 1 kHz | 20 to 30 dB down | Outdoor reflections, not a room. Slap rather than a tail |
| Passing car | 600 Hz to 1 kHz | rides up and down | The level move as it passes is what sells it, more than the filter |
This is the most common mistake and the most damaging. People pick a big hall preset and put it on the music. But the music is not in a hall, it is in a small room, and you are in a different small room. What you actually hear is filtered sound arriving in your space and bouncing around your walls. So the chain is: filter the music first, then feed it into a short reverb that matches the room the camera is in. Half a second, small, dry-ish. If the character is standing in a hallway, that reverb is a hallway.
Sound leaks through openings far more readily than it passes through material. A gap around a door can dominate the entire transmission path, and unlike the wall it leaks the whole frequency range, not just the bass. Practically: any time there is a door, a vent, a letterbox or a window that is even slightly open, blend a second copy of the music that is much less filtered and much quieter underneath the muffled one. That little bit of surviving detail is what stops it sounding like a blanket over a speaker.
Real walls leak. If you take everything above 800 Hz to silence, the ear hears a filter. Leave a trace, 30 or 40 dB down, and it hears a room. This is the same principle as the doorway above, and it is why a shelf often reads better than a hard low-pass.
A closed door is 25 to 30 dB, not silence. The number of scenes where the off-screen music is mixed far too quietly, then EQ'd to death to try to make it feel distant, is remarkable. Set the level first. Often you will find you need much less filtering than you expected.
Mute the picture and play only the treated music. If it sounds like a band in a room somewhere nearby, it is right. If it sounds like a band with a filter on it, the slope is too steep or the reverb is on the wrong side of the wall.
Lowpass, then Studio Reverb. Set the Lowpass cutoff from the table. Premiere's Lowpass is steeper than a real wall, so back the cutoff up a little higher than the figure suggests, or use the Parametric EQ with a high shelf instead for a gentler curve. In Studio Reverb, keep decay short and the wet level low.
Channel EQ for the shelf, then Space Designer for the room. Space Designer has small room impulses that do this job well, and because it is convolution the room reads as a real place rather than an algorithm. If you want to load third-party plugins here too, see using audio plugins in Final Cut Pro.
Fairlight's channel EQ plus the Reverb effect. Same order: filter, then space. Do it on a bus rather than the clip if the treatment carries across several cuts, otherwise you will be matching settings by hand every time you trim.
Any EQ and any short reverb. The technique matters more than the plugin. Put them on an aux so the sound of the space is one place you can adjust, not scattered over clips.
Everything above assumes a fixed listening position. As soon as the shot moves, the sound has to move with it, and that is where this stops being a five minute job.
A character walks down a hall towards a room where music is playing. As they approach, the cutoff opens up, the level rises, and the reverb changes from the hall to the room. Through the doorway it opens right out. That is three parameters moving together, in sync with picture, and doing it with keyframes means a lane of automation per parameter, redrawn every time the edit changes.
Most people give up and pick two or three static states, cross-fading between them at the cuts. That works, and if the cutting is quick nobody will notice. It falls apart on a long developing shot, which is exactly where you wanted the effect in the first place.
Rather than automating filters, you can describe where the sound is and where the listener is, and let the walls, floors, distance and doorway between them do the filtering. Move the listener and the filtering changes because the geometry changed, not because you drew it.
That is what we built THALM to do. You put the music in a room on a building cutaway, put the listening position somewhere else, and drag it around. Walk it through a wall and you hear the wall. There is a record mode: arm it, press play, drag the music through the building while the scene runs, and it keeps that pass as an editable route locked to the playhead. One take instead of a lane of keyframes.
It loads as an Audio Unit or VST3, so it goes straight into Final Cut, Premiere, Resolve, Pro Tools and any DAW. You can also drop your own impulse responses onto rooms if you have a library you trust.
No download and no signup for the browser version. It takes about thirty seconds to know whether it is useful to you.
We write these when there is something worth writing down. One email when a new one lands or a new Tunary instrument ships. No newsletter, no schedule.
Tunary