Make Jump Cuts Automatically
To make jump cuts automatically, cut the pauses out of a single continuous take. That operation is a jump-cut generator — removing time from inside one shot, with the camera in the same place on both sides, is the definition of a jump cut.
In Valmera you upload the take and type "cut all the pauses out". The agent finds the gaps in the speech from a word-level transcript, cuts each one at word boundaries so no syllable is clipped, and hands back a program with a junction wherever a gap used to be. There is no separate jump-cut button, and no editor anywhere has one, because there is nothing else a jump cut could be made of. The interesting half of this job is what happens next: which junctions you punch in on, which you cover, and which you leave alone.
Cut a Take into Jump Cuts
Upload one continuous take, type one sentence, review the junctions. 50 free credits, no card.
Try it free →How to Make Jump Cuts in Valmera
- 1Upload the takeDrag in the MP4, MOV, MKV or WebM — up to 3 hours. Valmera runs a one-time analysis: a word-level transcript with speaker labels, silence detection, shot detection, and labeled frame tiles the agent reads directly. It only happens once per upload.
- 2Type "cut all the pauses out"That single sentence produces the jump cuts. Defaults are a 0.5-second minimum gap and 0.12 seconds of padding kept around speech. Say "only cut pauses longer than a second" to make fewer, longer-lived junctions, or "cut tighter, leave less air" to make more.
- 3Punch in on alternating cutsType "punch in on the most emphasized lines" for an automatic pass, or name the moments: "punch in 15% from 0:42 to 0:44". A shot-size change across a junction is the accepted substitute for the camera move you cannot make in post.
- 4Cover the two or three that still snag, then exportWatch the preview once and mark the junctions where your hands or head teleport. Ask for coverage on those — "put the screen recording over the cut at 1:12 for two seconds". Then export: the render comes from your original file at source quality.
The analysis runs once per upload. Every step after it is a chat message, and any cut can be restored by asking.
- "cut all the pauses out"
- "cut the pauses and the umms, then punch in on my four most emphasized lines"
- "only cut pauses longer than a second — this is getting frantic"
- "punch in 15% from 0:42 to 0:44"
- "put the b-roll clip over the cut at 1:12 for two seconds"
- "restore the pause at 4:12, that one was deliberate"
What the Alternating Punch-In Is Doing
The picture below is the whole technique. Removing three pauses from one take produces three junctions where the shot is identical on both sides. Punching in on alternate spans means no junction has the same shot size either side of it — which is the 30-degree rule's own escape clause, bought in post instead of on set.
Why a Silence Pass Is a Jump-Cut Generator
The reason a jump cut is visible has nothing to do with the cut and everything to do with what the frames either side are carrying. At 30 fps, the last frame before the junction and the first frame after are a thirtieth of a second apart on screen and one to three seconds apart in real time. In that gap you breathed, shifted weight, moved your hands out of frame and turned your head. The picture is continuous in every dimension the eye uses to establish that this is one shot — background, lighting, lens, colour — and discontinuous in the one dimension that was moving. That mismatch is the jolt.
Classical continuity editing avoids the setup entirely. The 30-degree rule says two successive shots of the same subject must sit at least 30 degrees apart around it, and a significant change in shot size is the accepted alternative. Both remedies need a second camera position that a locked-off webcam does not have. Online video responded by breaking the rule on purpose: since mid-2000s vlogging made the style ubiquitous, audiences read jump cuts as pace rather than as error. What still reads as a fault is a cut that lands mid-word or clips a breath — a technical artefact, not a stylistic one, which is why cutting on the transcript rather than the waveform matters more than any effect you put on top. The full theory, including the spatial-versus-temporal distinction, lives in the jump cut glossary entry.
How the Cuts Are Actually Made
Valmera does not look for dips in the waveform. It looks for gaps in the speech, using the word-level transcript built during analysis — which is why a take recorded over a music bed, in a noisy room, or with game audio underneath still gets cut correctly, where a threshold-based detector finds nothing below the noise floor and reports that the video is already tight. Each detected gap is trimmed with 0.12 seconds of padding kept around speech by default, and only gaps of at least 0.5 seconds are considered. Every resulting cut is then snapped to a word boundary, so a junction can never land inside a syllable.
The cuts are entries in an edit decision list, not changes to your file. Your upload is never modified, so "restore the pause at 4:12" works at any point, even after twenty more edits. Previews render from a fast proxy; the final export renders from the original at source quality.
Two behaviours are worth knowing because they are the difference between an honest edit and a surprising one. If a removed gap was not actually quiet — nobody talking, but music or room tone playing — the agent is told to say so rather than call it a silence, because you are about to notice that audio jump. And every reply is verified server-side against the edits recorded, so the agent cannot report cuts it did not make.
The Four Things You Can Do With a Junction
1. Punch in — change the shot size
Ask for an automatic pass — "punch in on my most emphasized lines" — and Valmera scores the kept words by vocal stress and places three zooms at 14 percent by default, each running 0.9 seconds and starting 60 milliseconds early so the push lands on the word's attack rather than after it. Picks are spaced at least four seconds apart in program time, and any word the cut removed is skipped rather than estimated. You can ask for up to eight, at any strength from 5 percent up to 4.5 — a 5.5x blow-up, far past anything a talking head wants. The honest limitation: this pass aims at emphasis, not at junctions, so for a punch-in on one specific jump cut, name that moment — zooms can be placed over any range of the finished program and aimed at any point in the frame.
2. Cover it — b-roll or a cutaway
The only fix that removes the discontinuity rather than masking it, because the two frames the viewer would have compared are never adjacent. Insert a b-roll clip, a screen recording, a product shot or a generated still at the junction and give it 1.5 to 3 seconds — shorter reads as a flicker, longer becomes a scene. The insert can come from your own upload, a pasted link, stock, or AI generation, and it can be muted so your voice keeps running underneath.
3. Put a transition on it — sparingly
Seven styles are available, and by default they are placed only at real scene changes, decided by shot detection rather than by counting junctions. If every junction in your edit is a jump cut inside one continuous shot, Valmera declines to apply the transition at all and explains why, offering the explicit every-cut option instead. It also reports cadence: if the result would fire a full-screen effect more often than once every five seconds, it says so. Both behaviours exist because a preview once shipped with 45 whip pans through a single unbroken shot.
4. Leave it
Correct more often than the other three combined. Most junctions in a silence-cut take come from half-second pauses where you barely moved, and they are already invisible. Fix the four or five that snag; leave the other forty alone.
When It Goes Wrong
The pacing feels frantic. Too many junctions, not badly made ones. Raise the threshold — "only cut pauses longer than a second" — and the count drops sharply, because short gaps massively outnumber long ones in natural speech. Do not fix this with effects.
Every cut sounds like a clipped breath. This is an audio artefact being blamed on the picture. Increase the padding so more air survives around each sentence, and re-cut before you reach for punch-ins — a visual fix will not touch it.
Music or room tone jumps at every junction. The gap was not silent; it just had nobody talking in it. The agent flags this rather than calling those gaps silences. The fix is usually to cut fewer of them, or to lay the music on afterwards so it runs continuously over the finished cut instead of being chopped with the take.
The tight shots look softer than the wide ones. You have stacked too many punch-ins on too small a master. Reduce the strength, cut the number, or work from 4K.
Someone cut mid-word. Not possible here — cuts snap to word boundaries — but it is the characteristic failure of threshold-based tools, which decide where to cut from amplitude alone and have no idea where a word starts. If you have inherited an edit like that, re-cut from the transcript rather than patching the junctions one at a time.
How People Make Jump Cuts Without Valmera
Premiere Pro does this in Text-Based Editing: transcribe the clip, find the pauses in the transcript, and delete them — either lifting or extracting, with extract rippling the timeline closed. It is fast and it is built in. Read the transcript before you commit a bulk pause deletion rather than after, since the pass is only as good as the transcription under it. Punch-ins are then Scale keyframes on each clip, done by hand.
DaVinci Resolve transcribes audio on its Neural Engine and can ripple-delete the silent gaps straight from the transcription panel. Resolve is also the one mainstream tool that attacks the jolt directly: its Smooth Cut transition interpolates across a junction to smooth a jump cut in an interview. When it works on a locked-off talking head it is genuinely impressive; on large pose changes it warps. Note that the Neural Engine — and so the transcription those silence tools depend on — is a DaVinci Resolve Studio feature, not a free-version one.
ffmpeg is the free path, and it is real work. silencedetect prints silence start and end times to stderr; you parse those into a keep list, then build a filter graph of trim and atrim segments joined with concat. The trap is silenceremove, which strips silence from the audio only — apply it to a video and you have desynced the picture from the sound. There is no transcript involved at any point, so cuts land wherever the amplitude threshold says and mid-word cuts are on you.
The online silence removers and Premiere plugins — the dedicated category — mostly do the first half well, with threshold and padding controls and, in the better ones, a transcript. They stop at the cuts. Punch-ins, coverage and transition judgement are back in your editor. That division is the honest summary of the whole field: making jump cuts is close to solved everywhere, and deciding what to do with forty of them is not.
Honest Limits
Valmera does not synthesize frames. There is no equivalent of Smooth Cut here, and no AI interpolation across a junction — if you want the speaker's head to glide across a removed gap, Resolve's Smooth Cut is the tool. Valmera masks jump cuts; it never fabricates footage the camera did not record.
There is no true crossfade or dissolve. All seven transition styles are duration-preserving junction effects, and one style applies to the whole video — you cannot pick a different transition per cut. There is no motion tracking, so an insert placed over a junction cannot follow a moving subject. And there is no way to remove pauses without producing jump cuts — not here and not anywhere, because that is what the operation is.
One more, specific to the automatic punch-in pass: it needs speech. On a video with no transcript — silent screen capture, a montage of clips — it refuses rather than guessing, and tells you to place zooms by hand instead.
Related Tools
The same chat that makes the cuts can remove silence, strip filler words like um and uh, drop repeated takes and keep your best one, add word-accurate captions timed against the edit rather than the raw take, and place the punch-ins. One request can chain several: "cut the pauses and the umms, punch in on my four most emphasized lines, then caption it".
For the theory, read the jump cut entry and its neighbours on shot detection and J-cuts and L-cuts — which are unrelated to jump cuts despite the name. For the mechanics of cutting and restoring, see the cuts and cleanup docs, or browse everything in the tools hub.
Make the Cuts, Then Decide
One sentence produces the jump cuts. The next few decide how they read.
Start free →Frequently Asked Questions
Edit a Take into Jump Cuts
Upload real footage, describe the edit, review the preview, export at source quality. 50 free credits, no credit card.
Start editing free →