Cut silences and tighten pacing
Close dead air between clips and strip filler words with one prompt. Two tools, two jobs — know which one you're asking for.
Dead air kills retention. Instead of nudging clips together one by one, tell the agent to tighten the edit — it closes the gaps, ripples every track to keep sync, and you review the result. Two different tools handle two different kinds of dead air, and the agent picks the right one from your prompt.
Prompts to paste
"Tighten this edit — close the gaps and cut the filler words."
"Cut the silences between clips."
"Remove the ums, uhs, and repeated sentences."
"Close any gap longer than half a second, but keep the natural pauses."
"Caption this, then use the transcript to cut fillers and tighten pacing."
What the agent actually does
There are two tools here, and they work very differently:
Closing gaps. The agent scans the main video track for empty space between adjacent clips and shifts later clips left to close any gap longer than 0.3 seconds by default. It does not analyze audio at all — it never "listens" for silence. If your footage is one continuous clip, there are no gaps to close and this does nothing.
Cutting dead speech inside clips. Using the cached transcript (created when the agent captions your video), it removes filler-only segments ("um", "uh", "you know") and near-duplicate repeated segments, then ripples the main, overlay, and audio tracks to keep everything in sync. Timing is segment-level, not word-level, with a small trim padding (0.03s default) around each cut.
A prompt like "tighten this edit" typically triggers both: caption first if needed, then fillers, then gaps.
When to do it manually
Drag clips together on the timeline, or select a range and delete with ripple on — see Timeline editing. For surgical single-word cuts, split (S) at the playhead and delete the slice.
Limits
- Gap-closing only affects space between clips. Pauses inside one continuous clip are untouched — ask for filler removal (transcript path) instead.
- Filler removal needs a transcript first. If none exists, the agent captions the scene before cutting.
- Transcription is segment-level, so a cut lands on segment boundaries, not exact word edges. Very tight cuts can clip a breath — undo (Cmd/Ctrl+Z) and ask for more padding.
- Every change lands on the real timeline and is undoable.
See also
- Transcript highlights — pull the best moments, not just cut the worst
- Auto-caption a video — creates the transcript filler removal needs
- Cut by timestamps — precise range-based cutting with clean audio
- Timeline editing — the manual path