Addaly is in open beta. Things will change, and AI answers can be wrong — check anything that matters.

Editing Video

Cuts, sound, colour and export — taught on DaVinci Resolve, which costs nothing.

Lesson 8 of 779 min

Two cameras, one microphone, one sync point

Why sound and picture arrive separately

Camera microphones are close to the camera and far from the speaker, which is the wrong way round. So the good recording usually comes from somewhere else — a lavalier into a phone or recorder in a pocket, a shotgun on a boom into a field recorder, a second camera in another position.

That gives you two or more independent recordings of the same event, each started at a different moment, each running on its own quartz clock. Your job at the start of the edit is to make them agree.

The clap, and why it is still the best method

A single sharp sound, in frame, at the top of every take. A clapperboard if you have one, two hands if you do not.

It works because it produces one unmistakable vertical spike in every waveform and one identifiable frame in every picture. You line the spikes up and you are done. It takes fifteen seconds on set and saves half an hour per take in the edit.

Three details people get wrong:

  • Clap in shot, in focus, and audible on every device. A clap heard by the recorder but not the camera syncs nothing.
  • Clap at the start of the take. A clap at the end works, but you cannot use it until the take is over, and someone will stop recording early.
  • Clap again if anything stops. Every restart is a new sync problem.

Waveform sync, and when it fails

Every modern editor can align clips by comparing their audio waveforms. Resolve does it from the timeline or the media pool; Premiere has Synchronize; PluralEyes does it as a paid specialist.

It works by cross-correlation: sliding one waveform against the other and finding the offset with the highest match. It is accurate to the sample when it works, and it fails in four situations that are all common.

  • No shared sound. A camera 30 metres away in wind shares almost nothing with a lavalier.
  • Very quiet or very uniform audio. Steady room hum correlates with itself at every offset, so the peak is ambiguous.
  • Non-overlapping recordings. If the recorder ran continuously and the camera recorded four separate takes, the tool has to find four separate offsets in one long file — most will, some will not.
  • Different sample rates or variable frame rate. Which is the next section.

The workflow that works: try automatic sync, then check every take by eye at a hard consonant. A t or k sound has a visible attack, and a one-frame error is visible on a close-up of a mouth.

Drift: when sync is right at the start and wrong at the end

You sync the clap perfectly. Ten minutes later the lips are half a second ahead. Nothing slipped; the two devices simply counted time at slightly different rates.

Two causes, with different fixes.

Clock drift. Two independent crystals are never exactly equal. Over ten minutes a cheap recorder can gain or lose a few frames. Fix: stretch one clip's speed by the tiny ratio needed — if the audio is 4 frames late after 10 minutes at 25 fps, that is 4/15000, so set the audio clip's speed to about 99.973%. Resolve lets you type a duration and conforms the clip to it.

Variable frame rate. Phones and screen recorders often write VFR: the nominal rate is 30 fps but the actual rate wanders with light and processing load. The editor assumes a constant rate, so picture and sound diverge steadily. This is by far the more common cause with phone footage. Fix it before importing, by transcoding to constant frame rate:

Two causes of drift, across one ten-minute take01020010Minutes into the takeFrames out of sync at 25 fps—— Phone file at variable frame rate– – Cheap recorder, clock drift– – Both conformed at offloadThe clap was right. Thirteen frames is half a second of lip sync. Transcode the phone file to constantrate at offload and the top line becomes the bottom one.
Two causes of drift, across one ten-minutetake01020010Across: Minutes into the takeUp: Frames out of sync at 25 fps—— Phone file at variable frame rate– – Cheap recorder, clock drift– – Both conformed at offloadThe clap was right. Thirteen frames is half a secondof lip sync. Transcode the phone file to constantrate at offload and the top line becomes the bottomone.
bash
ffmpeg -i phone.mp4 -vsync cfr -r 30 -c:v libx264 -crf 18 \
       -c:a aac -b:a 320k phone_cfr.mp4

Do this to the whole shoot at offload and the problem disappears for the rest of the project.

Timecode, for when there are more than two devices

Professional sets solve all of this with timecode: every device stamps each frame with a time of day from a shared reference, and the editor aligns by number rather than by sound. Timecode generators — Tentacle Sync and similar — cost real money.

The free approximation: set every device's internal clock to the same second before the shoot, and use file creation time as a coarse alignment before waveform sync refines it. It gets you into the right minute, which is enough for the automatic tools to find the right second.

Multicam, practically

Once clips are in sync, group them into a multicam clip — Resolve calls it a Multicam Clip, Premiere a Multi-Camera Source Sequence, Kdenlive does it through its own mechanism. You then play the timeline and cut between angles live by pressing 1, 2, 3, and refine afterwards.

Two habits make the result watchable rather than frantic. Set the audio to come from one source — the good microphone — for the whole multicam clip, so switching angles never switches sound. And give each angle at least three seconds before switching away, unless the cut is on an action; a multicam cut every second reads as a person who could not decide.

Today

Record ten seconds on your phone and ten on any second device, clap once at the start, and sync them by hand. Then let the automatic tool do it and compare. You will trust it appropriately after that.

The one thing to keep

Give every take one sharp sync point at the start, use waveform sync as a first pass and check it at a hard consonant, and convert variable-frame-rate phone footage to constant rate at offload because drift is usually the file, not the clocks.

Before you move on

An interview is shot on a phone with a separate audio recorder. The clap syncs perfectly, but by minute eight the lips lead the sound by about half a second. Which explanation should be checked first, and why?

Pick the one you would defend. Nobody sees your answer.

No ads. No data sale. No public scores on people. Ever.

© 2026 Addaly