Cut the Sequence on the Sound

Cut on the sound, not on the picture. Find the breath before a line and the transient at the start of a word, place the cut one or two frames ahead of the visual change, and let the audio lead. Picture cuts that land on a hard consonant or mid-breath read as mistakes even when the frame is perfect.

This is the method the studio uses on every dialogue sequence, and it costs nothing but a second pass over the waveform. Editors who cut picture first and fix audio later spend that time twice.

Why does the sound decide where the cut goes?

Because the ear resolves timing finer than the eye. A viewer tolerates a picture cut that is three frames early; the same viewer hears a breath chopped in half. Speech carries its own punctuation: the inhale before a sentence, the release after it, the small stop where a speaker changes thought. Those are the joints. Cut there and the sequence breathes with the person on screen.

Work in this order. Lay the audio of both takes end to end, no picture trimming yet. Read the waveform at high zoom, one second across the screen. Mark the breath, mark the transient. Then cut picture to those marks, not the reverse. The same discipline shows up outside editing: practitioners writing about creative work as a resource for care describe attention to a person's own rhythm before imposing a structure, which is the same instruction in another room, and you can read more of that thinking at art that heals.

Cut on the breath and the transient

A transient is the first 5 to 20 milliseconds of a sound, the click that tells the ear a new event started. A breath is the 200 to 600 milliseconds of air before a phrase. Both are visible in any editor's waveform display, and both are better cut points than the frame where a mouth opens.

Practical placement:

• Dialogue, same speaker, continuous thought: cut 2 frames before the transient of the incoming word. The ear hears the new word start clean.
• Dialogue, speaker change: cut on the outgoing speaker's breath, 1 to 2 frames after the last consonant. The listener gets the pause as information.
• Reaction shots: cut 4 to 6 frames after the line ends, so the face lands while the room tone is still ringing.
• Room tone under everything: keep one continuous bed across the whole scene, 3 to 6 dB under dialogue. Cut picture over it, never the bed itself.

If a cut feels wrong and the frame looks right, move it 2 frames in either direction and listen again. Nine times out of ten the problem is the transient, not the composition.

What are J cuts and L cuts for?

They buy time without spending picture. A J cut brings the next scene's audio in before its picture: the listener hears the new room, then sees it. An L cut holds the previous scene's audio over the new picture: the listener stays in the old room while looking at the new one.

Use them at scene boundaries where a hard cut would feel like a door slamming:

• J cut, 12 to 24 frames of audio lead. Enough to establish a new space, short enough that the viewer does not wonder where the picture went.
• L cut, 8 to 16 frames of audio tail. Good for walking away from a subject while their sentence finishes.
• Both at once, offset by 1 to 2 seconds, when you want a dissolve in sound without a dissolve in picture.

Overlap is the same idea inside a scene. Let the outgoing line's last syllable run 3 to 5 frames under the incoming shot. The two shots stop feeling like separate clips and start feeling like one continuous conversation.

How do you keep sync when the picture is under a music bed?

You keep sync by trusting the waveform, not the beat. A music bed tempts you to cut picture on the downbeat, and that is fine for a montage, but it destroys lip sync in dialogue. Two rules cover most cases.

First, never move a synced shot to hit a musical accent. Move the music instead. Nudge the bed 1 to 3 frames, or cut a bar, so the accent lands where the picture already is. Music is elastic; a mouth is not.

Second, check sync at three points per shot: the first frame, the middle, and the last. Play at quarter speed with audio at full level. If the plosive of a P or B arrives before the lips close, the shot is out by 1 to 2 frames. Fix it in the timeline, not with a global offset, because the drift is usually per shot.

When the bed sits under dialogue, duck it 6 to 9 dB with a 150 to 250 millisecond release. A hard duck sounds like a gate opening; a slow one lets the music swell back over the next line. If the bed has a strong 2 to 4 kHz presence, cut 2 to 3 dB there with a wide bell before ducking, and you will need less level reduction overall.

Build the sequence in passes

Three passes, in this order, and the sequence holds.

1. Audio assembly. All takes, all room tone, no picture trimming. Mark breaths and transients with markers.
2. Picture cut to the marks. Ignore music entirely. Watch the scene silent once, then with dialogue only.
3. Music and overlap. Add the bed, set the J and L cuts, check sync at quarter speed.

Skipping pass one is the usual failure. Editors who trim picture first end up with cuts that sit on the wrong side of a breath, and no amount of music hides it. The fix costs a full rebuild.

What goes wrong when the sound is an afterthought

Four failures show up again and again.

• Breath cut in half. The listener hears a gasp. Move the cut 2 frames later.
• Transient clipped. The first consonant disappears under the previous shot's room tone. Extend the incoming audio 3 frames earlier.
• Sync drift under music. The bed was cut to fit the picture instead of the picture being checked against the bed. Recheck at quarter speed.
• Overlap too long. More than 8 frames of audio tail and the two speakers start talking over each other. Trim to 3 to 5 frames.

None of these need new gear. They need the waveform on screen at one second across, and the discipline to cut where the sound says.

Sources

• Walter Murch, In the Blink of an Eye, on the rule of two and a half frames and cutting on emotion before story, rhythm, and eye trace.
• BBC Academy, editing and post-production guides, on J and L cuts and dialogue assembly.
• Acoustics and psychoacoustics references on transient perception and the precedence effect, for the timing thresholds quoted above.