J-Cuts and L-Cuts (Split Edits)

J-cuts and L-cuts are split edits: cuts where the sound and the picture change at different moments. In a J-cut, the audio of the next shot starts before its picture, so the audience hears what is coming before they see it. In an L-cut, the audio of the current shot carries on after the picture has cut away. The names come from the shape the two clips make on a timeline where audio sits under video: the incoming clip's audio reaches back under the previous picture like the bottom of a J, and the outgoing clip's audio trails forward like the foot of an L. Split edits are one of the oldest tools in filmmaking, and they have become a test of whether automated editing tools understand conversation or only chop it.

Why Split Edits Feel Natural

A straight cut, where sound and picture change on the same frame, draws attention to itself in dialogue. Real attention doesn't work that way. In a conversation, people often hear someone begin to speak and only then turn to look, which is exactly what a J-cut reproduces. And people keep watching a speaker's face after they have heard the end of a sentence, or glance at the listener while the speaker is still talking, which is what an L-cut reproduces. Split edits therefore do three jobs:

• Hide the cut. When the sound is continuous across a picture change, the viewer follows the sound and barely notices the edit.
• Show reactions. An L-cut lets the viewer see a listener's face while the speaker finishes, which is often where the emotional information is.
• Bridge scenes. A J-cut can introduce a new place or time with sound (a crowd, a train, a voice) before the picture arrives, so a scene change feels motivated rather than abrupt.

In interviews and documentaries, L-cuts also let an editor cover a jump in a speaker's answer with B-roll or a cutaway while the voice runs on, which is how a long answer is tightened without the picture visibly jumping.

How They Are Made

Every professional editor, including Avid Media Composer, Adobe Premiere, DaVinci Resolve and Final Cut Pro, supports split edits, typically by unlinking a clip's audio from its video or by trimming the audio edge separately from the video edge. A J-cut is sometimes called an audio lead or audio advance. The craft is in the timing: a lead of a few frames is barely perceptible and simply smooths the cut, while a lead of a second or more becomes a deliberate storytelling choice.

Split edits also interact with the mix. When two clips overlap in sound, the editor has to decide whether the voices cross-fade, whether room tone matches, and whether the incoming line starts during a pause. A J-cut that starts in the middle of the previous speaker's word sounds like a mistake, not a style.

Split Edits in Automated Cuts

Automated editing, from multicam auto-switching to agentic video editing, has to make the same decision a human editor makes: when to change the picture relative to the sound. The naive approach switches cameras the instant a different person starts talking, which produces a cut on every first syllable and a stuttering, mechanical rhythm. Better systems use speech recognition word timestamps and speaker diarization to find speech boundaries, then place the picture change inside a pause and let the sound lead or trail it. Professional editors now ship this. DaVinci Resolve 20 (May 2025) added AI Multicam SmartSwitch, which switches multicam angles to the active speaker. Avid Media Composer 2026.8 (August 2026) added PhraseFindAI Multicam Auto Cut: the editor assigns speakers to camera angles in the transcript tool, and Media Composer builds an editable first-pass multicam edit that switches angles by who is speaking and cuts on speech boundaries. Such a first pass is a starting point; the editor still decides where a reaction shot or a held line works better than a switch.

For an AI system, a split edit is also a constraint to check. If a clip's source doesn't contain enough sound before its in-point, a requested lead can't be honored in full; if a split cuts off the end of a word, the result is audible. Systems that treat an edit as data, as in video as code, can verify these conditions before rendering instead of discovering them in review, and can hand the result to a human editor through timeline interchange formats that preserve the separate audio and video edit points.

Common Mistakes

• Overuse. Splitting every cut makes a conversation feel slippery; many cuts should still be straight.
• Clipped words. The audio edge lands mid-word, either at the start of a J-cut or the end of an L-cut.
• Two voices at once. An overlap long enough that both speakers are heard together, which reads as an error unless the people really did talk over each other.
• Mismatched room tone. The background sound changes on the picture cut, revealing the edit the split was meant to hide; ducking and tone fills are the usual fixes.