Audio Post-Production: Dialogue, ADR and Foley for Video.

A video is judged by what is heard almost before what is seen. Dialogue requiring deciphering, a jumping voice edit or footsteps missing as a character walks are enough to disengage the listener, however good the picture. Audio post-production is the set of tasks turning raw location sound into a soundtrack that holds up on its own. Most of the work is editing and judgement rather than effects.

This chapter covers Audio Post-Production, Dialogue, ADR and Foley: the workflow connecting the four, what to do in each and how to avoid a later correction forcing you to redo work that was already right.

What Audio Post-Production Is

Audio post-production begins when the reference picture and sound recorded during filming arrive, usually as a session exchange with linked media. The usual route is then to review the whole take, discard or mute unusable material, mark what needs replacing, edit usable material, add missing elements and mix for delivery.

The first hour decides the result. Opening the picture, listening to production sound from beginning to end and noting problems, such as noise-covered lines, shots with distant voices and script changes, avoids discovering those same issues halfway through a session. Everything unfixable is marked as an ADR or Foley-recreation candidate; everything fixable is left for editing.

Also separate two often-confused activities: content editing and sound editing. The first changes when things are said, by trimming a passage, removing a filler word or ordering takes, and affects picture synchronisation. The second changes only timbre and level, leaving timing untouched. Doing them in the wrong order, processing the voice and then moving clips, forces you to repeat processing because each movement changes the relationship between the take and the background noise you cleaned.

Settle two project decisions before starting: the sample rate, set to 48 kHz for video to avoid conversion problems, and the synchronisation reference to picture, covered in detail in the guide to audio and video synchronisation with timecode. Once those are fixed, the session stops shifting and the work becomes linear.

Dialogue: Editing without Revealing the Edit

Dialogue is the most listened-to layer and the least tolerant of careless editing. The aim is for nobody to notice any intervention: cuts should fall between phrases or in short breaths, with room tone before and after each cut.

The working order is gain first, processing afterwards. Match takes with clip gain, raising quiet parts without jumps, and only then apply track EQ or compression. Reverse that order, and each correction forces you to redo the previous one.

Cleaning has two fronts: background noise and room reverberation. Spectral denoisers remove hum and rubbing without harming the voice, and reverberation reducers help when the scene changes location. Both produce artificial-sounding artefacts if overused: a voice with clean sibilance but its air stripped away reveals more than mild noise.

Finally, room tone. Recording thirty seconds of silence at each location lets you join lines, extend pauses and fill gaps without audible seams. Without room tone, every cut becomes a small explosion of digital silence.

ADR: When It Is Needed and How It Is Recorded

ADR (Automated Dialogue Replacement) replaces lines from the original take with ones recorded later in a studio, synchronised to picture. It is a resource rather than a routine: use it when production sound cannot be saved, because of machine noise, wind, overlapping lines or a technical fault, or when the line changes in the script.

The process is consistent. Prepare the script with marked lines, import video with the original track as a reference, place markers for each line and loop-record to gather variants without interrupting the performer. Then comp the best version take by take, remove excess silence, align timing to the reference and fade each clip’s edges to avoid clicks.

The mix determines whether ADR goes unnoticed: match tone to the original track, simulate the real space’s reverberation with an impulse response from that location and automate level so the line follows the scene. A new take with the wrong room is noticed sooner than an old take with noise.

Foley: Actions Supporting the Scene

Foley manually recreates sounds accompanying picture: footsteps, clothing movement, doors, handled objects and falls. It is recorded separately because location production sound includes equipment noise, overlapping dialogue or a perspective that does not fit the shot.

The method has not changed: review the scene on screen, identify actions needing sound, prepare surfaces and objects, and record with a directional microphone very close to the action. Always record with headphones to detect fans, the operator’s clothing rubbing or traffic, and record variants of each action because the editor almost always needs to choose and repeat.

Two habits separate convincing Foley from obvious Foley: distance, which should follow the camera perspective, and file coding by scene, location and character, so editing finds each item without searching twice. That is how you build your own action library, recording by recording.

For a small piece without a Foley room, a quiet room with varied surfaces works. The capture technique from recording voices and instruments in the studio applies equally: close, directional and with level headroom.

The Working Order That Avoids Rework

The four elements have dependencies, and breaking them costs hours. Dialogue is edited and stabilised first because it is the whole scene’s level reference. Foley actions then enter against picture with no further picture edits. Supporting effects and ambience follow, and only finally comes mixing, deciding balance and delivery level.

Two rules keep the session sound: work at 48 kHz and 24 bits throughout, and use the same naming convention for tracks and files, including scene, character and take type, from day one. When material comes from several sources, that convention alone prevents uncertainty over which version is correct.

Also decide early how far processing should go. Monitor levels from the start with the same criteria as the guide to gain staging and decibel meters, so the mix does not reach the end with its headroom already spent.

Mistakes Revealing a Rushed Delivery

Five faults explain most amateur-sounding deliveries. First: cutting dialogue without room tone, so joins sound like pulses. Second: recording ADR without matching the room and the original voice tone. Third: placing Foley at the wrong distance, with footsteps more prominent than the characters. Fourth: pushing everything to the limit and losing the space the picture needs to breathe. Fifth: delivering without checking synchronisation in the viewer, with the first and last takes out of place.

None requires more tools, just a complete listen before exporting, with the checklist in front of you and video full-screen.

Add a sixth, the most avoidable: exporting without reviewing the file’s first and last half-second. That is where clicking cuts, uncontrolled levels and remnants of another mix version appear. Repeat this check on headphones and speakers, and compare with a reference of similar volume. That is the difference between a file that passes and one returned for corrections.

Next Step

With dialogue stable, ADR resolved and actions recorded, complete the soundtrack: effects, ambience, music and delivery level. All of that is in the guide to sound effects, ambience, music editing and loudness for video.