Wedding Film Audio: How to Fix Sound in Post

The short answer: Fixing wedding film audio in post runs in a fixed order: sync every source, choose one source per moment, repair the damage, then balance and mix. Target roughly -16 LUFS integrated with true peaks below -1 dBTP for web delivery. Most bad wedding audio is not unfixable — it is unchosen, because the editor left three sources fighting each other underneath the same moment.

Search for wedding film audio and you will find the same article fifteen times. Put a lav on the groom. Get a feed from the DJ. Wear headphones. Watch your levels.

All true, all useful, and all about the day.

None of it helps on the Tuesday afterwards, when the lav rustled through the vows, the board feed clipped during the speeches, and the only clean take of the line that matters is buried under a photographer’s shutter.

This is about that Tuesday.

Why sound is the half that gets skipped

Audio carries the emotional content of a wedding film almost entirely. Vows, speeches, a laugh caught between two people — none of it survives the picture alone. Yet sound routinely receives a fraction of the attention given to colour, because colour is visible in a still frame and sound is not.

The practical evidence is easy to check on your own work. Mute a wedding film and watch it. It becomes a montage. Now play it with picture off and only audio. It usually still works.

A well-mixed film with average colour
beats a beautifully graded film with poor sound.

Every time. And unlike a grade, bad audio is not a stylistic choice a viewer can be talked into. It reads as a mistake.

The order: sync, choose, repair, balance

Wedding audio post-production follows four stages in sequence: sync all sources to picture, select a single source per moment, repair damage in that source, then balance it against music and mix to a delivery target. Working out of order is the most common cause of a mix that never quite settles.

The stage editors skip is the second one.

• Sync. Lavs, board feed, and camera scratch audio aligned to picture. Waveform sync handles most of it; check the ceremony manually because that is where drift shows.
• Choose. One source per moment. Not three at low volume underneath each other.
• Repair. Noise, dropouts, sibilance, clipping — fixed on the chosen source only.
• Balance. Dialogue against music, then the whole mix against a loudness target.

Most muddy wedding audio is not a repair failure. It is three sources running simultaneously because nobody decided which one was carrying the moment. Two lavs and a board feed of the same vow produce phase smear and a hollow, distant quality no plugin will fix — because the problem is arithmetic, not noise.

Choosing a source, moment by moment

The best source for a given moment is the one physically closest to the sound, not the one that is technically cleanest. A slightly noisy lav on the groom during vows will always beat a pristine board feed picking him up from twelve metres away, because proximity carries intimacy and no amount of processing adds it back.

What each source is actually for:

• Lav on the groom / officiant. The vows. Almost always the primary. Close, intimate, and the reason the ceremony works at all.
• Board or DJ feed. Speeches and anything through the venue PA. Clean signal, but it is a line feed — it has no room in it and can sound sterile alone.
• Camera scratch audio. Rarely the primary, always the safety net, and the best source of usable room tone.
• A dedicated recorder in a pocket. The backup that saves the film roughly once a season.

A useful habit: blend a very low level of camera or room audio under a board feed — around 15 to 20 dB below it. You are not adding intelligibility. You are adding the room back, so the speech stops sounding like a phone call.

The five problems that actually arrive

In practice the same handful of faults turn up on almost every wedding. These are the fixes, in the order they should be applied.

Lav rustle and clothing noise

Caused by fabric moving against the capsule. Fix with a narrow EQ cut rather than broadband noise reduction — the energy usually sits between 200 Hz and 1 kHz and is short. Broadband reduction applied to rustle takes the voice with it. If the rustle sits over a word that matters, cut to the board feed for that phrase and cut back.

Wind on outdoor ceremonies

High-pass filter at 80 to 100 Hz on speech removes most of it without touching intelligibility — there is nothing in a human voice below 80 Hz worth keeping. Steeper slopes sound unnatural; 12 dB per octave is usually enough.

Clipped speeches

A board feed pushed too hard clips, and clipping is genuinely destructive — the waveform information is gone. De-clip tools reconstruct plausibly at mild levels and sound artificial past that. Where clipping is severe, the honest answer is to use the camera or room source for that passage even though it is further away.

Room echo in stone churches and barns

Largely unfixable, and worth knowing that. Reverb reduction tools help slightly and cost you naturalness quickly. The better move is structural: use the close lav for the words and cover the reverberant wide audio with music underneath.

Dropouts on wireless

A moment of silence where a word should be. Patch from another synced source if one exists. If not, and it is a single syllable, a short crossfade from the surrounding room tone is less noticeable than the hole. Never leave true digital silence in a wedding film — it reads as a technical fault, where room tone reads as a pause.

Cleaning without sterilising

Noise reduction should be applied conservatively, typically no more than 6 to 10 dB of reduction, because aggressive processing produces the hollow, underwater quality that is more distracting than the original noise. Wedding audio is supposed to sound like a room containing people.

A workable chain, in order:

1. High-pass          80-100 Hz, 12 dB/oct    remove rumble and wind
2. Noise reduction    6-10 dB max             hiss, air conditioning
3. Subtractive EQ     narrow cuts only        boxiness around 300-500 Hz
4. De-esser           5-8 kHz, gentle         sibilance from close lavs
5. Compression        3:1, slow attack        even out level, keep dynamics
6. Gentle limiter     ceiling -1 dBTP         safety only, not loudness

Two rules that matter more than the settings. Cut before you boost — removing what is wrong beats adding what is missing, and it costs no headroom. And bypass the whole chain every few minutes. Ears adapt within seconds, and processing creep is how a mix ends up sounding processed to everyone except the person who made it.

Ducking with intent

Music should move under dialogue rather than simply becoming quieter. A typical duck of 12 to 18 dB under speech, with attack and release timed to the phrasing rather than the transient, keeps the music present without competing. Automatic sidechain ducking produces a mechanical pumping that draws attention to the mix itself.

The practical difference is where you place the movement. Automation ducks on the word. A person ducks on the breath before the word, and comes back up on the natural gap after the sentence — not the instant the mouth stops moving.

It is slower to do by hand. It is also the single most audible difference between a wedding film that sounds edited and one that sounds mixed.

Loudness targets for delivery

Wedding films for web delivery should be mixed to approximately -16 LUFS integrated with true peaks below -1 dBTP. Vimeo and YouTube both normalise loudness on playback, so mixing louder than the target gains nothing and costs dynamic range — the film simply gets turned down, with the transients already squashed.

DELIVERY               INTEGRATED      TRUE PEAK
Web / Vimeo / YouTube  -16 LUFS        -1 dBTP
Instagram / social     -14 LUFS        -1 dBTP
Broadcast (EBU R128)   -23 LUFS        -1 dBTP

DIALOGUE SHOULD SIT    -12 to -6 dB below peak music
MUSIC UNDER SPEECH     12 to 18 dB below its own level

Consistency across your deliverables matters more than any individual number. A couple watching a teaser, a highlight film and a feature back to back should never reach for the volume control between them.

The test that catches everything

Play the finished film at low volume, on laptop speakers, in a room with other noise in it.

If the vows still carry, the mix works. If you find yourself leaning in, the music is too loud — regardless of how balanced it sounded on headphones at midnight.

Almost nobody watches a wedding film the way it was mixed. They watch it on a phone in a kitchen, on a laptop in bed, on a television with a family talking over it. Mixing on good headphones and never checking anything else is the most common reason a film that sounded finished arrives sounding thin.

A second check worth thirty seconds: play it on your phone speaker. Phone speakers reproduce almost nothing below 300 Hz, so anything relying on low-end warmth disappears entirely. If the film still holds, it holds anywhere.

Where sound sits in the wider process

Sound is the fourth stage of six. It comes after the story and the cut are locked, and before the colour grade — because mixing to a moving edit means mixing twice, and grading a film whose sound is not settled means judging emotional weight that is not there yet.

The full sequence is in our complete guide to wedding video editing, and the colour stage that follows this one is in the guide to wedding film colour grading. If you are working out where the hours go, how long it takes to edit a wedding video puts sound at roughly 1 to 1.5 hours of a 7 to 10 hour highlight film.

That is a small share of the edit for something that carries half the result.

Wedding film audio: common questions

How do you fix bad audio in a wedding video?

Work in order: sync all sources, choose one source per moment rather than layering them, then repair that source with a high-pass filter, conservative noise reduction of 6 to 10 dB, gentle de-essing and light compression. Most poor wedding audio is caused by multiple sources playing simultaneously, not by damage that needs repairing.

What loudness should a wedding film be mixed to?

Around -16 LUFS integrated with true peaks below -1 dBTP for web and Vimeo delivery, or -14 LUFS for social platforms. Mixing louder gains nothing because both YouTube and Vimeo normalise playback loudness, so a hotter mix is simply turned down with its dynamics already reduced.

Should I use the lav mic or the DJ soundboard feed?

Use the source physically closest to the sound. The lav is almost always right for vows because proximity carries intimacy that processing cannot add. The board feed is better for speeches through the PA, though it benefits from a small amount of room audio underneath — around 15 to 20 dB down — so it does not sound sterile.

How do you remove wind noise from wedding video audio?

A high-pass filter at 80 to 100 Hz with a 12 dB per octave slope removes most wind rumble without affecting speech, because there is nothing in a human voice below 80 Hz worth keeping. Broadband noise reduction is the wrong tool and takes the voice with it.

Can clipped wedding speech audio be fixed?

Only partially. Clipping destroys waveform information rather than masking it, so de-clipping tools reconstruct plausibly at mild levels and sound artificial beyond that. Where clipping is severe, a more distant camera or room source usually sounds better than a heavily repaired clipped one.

How much should music duck under speech in a wedding film?

Typically 12 to 18 dB below its own level, automated by hand rather than by sidechain. Duck on the breath before the line and return on the natural gap after the sentence. Automatic ducking triggers on the transient and produces a mechanical pumping that draws attention to the mix.