Cuts and cleanup
Ask Valmera to cut silences, filler words, retakes or any passage you describe, and it edits from the word-level transcript: cuts snap to word edges so no word is clipped, and every cut can be restored. Name passages by their words (“cut the part about the delay”) or by timestamp.
Remove silences and dead air
Ask “cut the silences” and Valmera removes every pause of 0.5 seconds or longer, keeps about 0.12 seconds of breathing room around speech, and snaps each cut to a word edge. It handles the whole video in one pass, then checks the kept transcript before it reports back.
- Open the analyzed project
Cleanup works from the word-level transcript, so wait until analysis finishes (or type your request and let it start automatically).
- Ask for the cleanup
Type “Cut the silences and remove the ums.” Add what must stay, for example “keep the pause before the punchline.”
- Review the preview
Listen at several cuts: the first and last sound of each word should be intact and the rhythm should feel natural.
- Tune it in plain words
Ask for more or less breathing room, a different minimum pause, or restore any passage that should come back.
- Cut the silences and remove the ums.
- Tighten the pacing, but leave a little more breathing room between sentences.
- Only cut pauses longer than one second.
- Cut the dead air, but keep the pause before the punchline at 2:14.
Type these in Studio chat, or ask Claude, ChatGPT, Cursor or another assistant connected to Valmera’s MCP server. More in the prompt library.
Remove filler words
“Remove the ums” cuts hesitation sounds such as um, uh, er and hmm using exact word timestamps. Words that can carry meaning, like like, you know, so or basically, are only removed when you name them.
- Remove the ums and uhs.
- Also remove “you know” and “basically”.
- Remove filler words from the intro only, up to 0:45.
Type these in Studio chat, or ask Claude, ChatGPT, Cursor or another assistant connected to Valmera’s MCP server. More in the prompt library.
Remove retakes and repeated lines
Recorded the same line three times? Valmera finds the repeated attempts in the transcript and keeps one. Tell it which take to keep, or let it choose the complete one and check the preview. After cutting repetitions, it reads the kept transcript again to confirm no repeated phrase survived.
- Remove the false starts and keep the last full take of each line.
- In the intro, keep the take that starts with “Hi, I’m Sam” and cut the others.
- I repeated the pricing explanation. Keep the second version.
Type these in Studio chat, or ask Claude, ChatGPT, Cursor or another assistant connected to Valmera’s MCP server. More in the prompt library.
Cut any passage
Describe what to remove by its words, its topic or its time. Valmera finds it with a transcript search and cuts on word boundaries. Cutting inside an inserted clip works too: the clip is split around the cut.
- Cut the section where I talk about the delayed shipment.
- Cut 1:05 to 1:20 of the current video.
- Keep only the answer to the second question.
- Trim the first five seconds and the last ten.
Type these in Studio chat, or ask Claude, ChatGPT, Cursor or another assistant connected to Valmera’s MCP server. More in the prompt library.
Source time and video time
Timestamps in the player and timeline are video time: positions in the edited video. Timestamps in the transcript are source time: positions in your original recording. After cuts they differ, so say which you mean when you give numbers: “cut 0:12–0:15 of the edited video” or “cut source 3:40–3:52.”
To trim by hand, drag the edge of a footage block on the timeline; dragging it back restores the footage. See edit the timeline.
Restore anything
Cuts never delete footage. Ask “restore the part about pricing” or “put back source 2:10–2:25” and only that passage returns; everything else stays as it is. To undo a whole request, step back with Previous edit, or ask “go back to the version before the last change.” To begin again, ask to start over from the full source: earlier versions stay available.
- Restore the sentence that starts with “The real reason…”.
- Put back the last 10 seconds I cut from the ending.
Options and defaults
| Setting | Default | How to change it |
|---|---|---|
| Minimum pause to cut | 0.5 seconds | “Only cut pauses longer than 1 second” |
| Breathing room around speech | About 0.12 seconds | “Leave more room between sentences” |
| Word edges | Cuts keep whole words | Name the exact moment if you need a hard cut |
| Filler words | um, uh, er, hmm and similar | “Also remove ‘like’ and ‘you know’” |
| Pauses that contain sound | Protected when the cut would be large | “Cut them anyway” |
| Scope | Whole video | “Only in the first two minutes” |
Troubleshooting
| Symptom | Fix |
|---|---|
| A word sounds clipped at a cut | Ask to widen that join: “Give the cut before ‘so the next step’ a quarter-second more room.” |
| The pacing feels rushed | Ask for more breathing room, or a longer minimum pause (for example 0.8 seconds). |
| “No filler words were found” | The transcript has no hesitation sounds to cut. Name the words you hear: “remove ‘so’ at the start of sentences.” |
| The silence pass stopped and asked first | Many gaps contain music or action. Reply “cut them anyway”, or name the pauses to remove. |
| The wrong part was cut after you gave a time | Say whether you meant video time or source time, or quote the words instead. |
| The wrong take was kept | Quote the take to keep: “keep the one that starts with ‘Okay, so…’.” |
| Room tone jumps at cuts | Add a quiet music bed under the speech, which smooths small changes in background sound. |
Limits
- Cleanup reads the transcript, so it needs speech in the main video; speech inside inserted clips isn’t transcribed.
- Removing silence doesn’t remove background noise. Valmera has no audio denoise.
- English is the best-tested transcription language; check cuts closely in other languages.
Tools behind this page: cut_silences find_silences remove_filler_words cut_range cut_output_range keep_segments restore_range search_transcript get_kept_transcript reset_edit · Full toolset reference
Frequently asked questions
Will silence removal clip my words?
Cuts snap to word edges and keep about 0.12 seconds of breathing room around speech, so words stay whole. If one join still sounds tight, ask to widen it, for example “give the cut before ‘so the next step’ more room.”
Can Valmera remove “like” and “you know”?
Yes. By default filler removal cuts hesitation sounds such as um, uh, er and hmm. Name any other words or phrases (“also remove ‘like’ and ‘you know’”) and Valmera targets exactly those. Review the sentences where they were cut, because those words can carry meaning.
Can I undo a cut?
Yes. Cuts are non-destructive. Ask to restore a passage by its words or time and only that passage returns, or step back to an earlier version with Previous edit.
Does it work on podcasts with two or more speakers?
Yes. The transcript has speaker labels, so you can ask to keep each speaker's answers complete, cut one person's tangents, or keep only the guest's answers.
What's the difference between source time and video time?
Source time is a moment in your original recording; video time is a moment in the current edited video. After cuts they differ. When you give a timestamp, say which one you mean, for example “cut 1:05–1:20 of the edited video.”
Can I cut by describing the topic instead of a timestamp?
Yes. “Cut the part where I talk about the delayed shipment” works: Valmera searches the transcript for the passage and cuts it on word boundaries.