Updated November 9, 2026. Desktop workflow; controls and capabilities depend on the app version.
Decide what should actually be removed
To remove filler words from video, identify unwanted speech, cut its matching interval and check the join. A transcript can help you locate “um” or “uh,” but an automatic list is only a set of candidates.
Keep words that carry meaning. “I actually tested this” can be a useful contrast; “like this” can point to an on-screen action. Pauses may also be necessary for a demonstration, emphasis or a speaker’s natural rhythm. There is no fixed amount of every recording that should be cut.
Generate captions before finding spoken fillers
Open your recorded or imported video in Penbeam desktop and choose Generate subtitles. The first use requires a speech-model download; later transcription runs locally with the prepared model. Check that the source has an audible speech track, and proofread the transcript.
Filler detection depends on recognised words and timing. If “um” was omitted by transcription, it may not be detected automatically. A multi-word phrase may also need manual selection. You can still listen and select the unwanted interval in the transcript or timeline rather than treating a missing detection as proof that the recording is clean.
Use the Script view for review-first removal
- Open Text edit → Script after captions have been generated.
- Use Analyze to populate the panel with speech, pause and filler candidates.
- Listen to each questionable part. Use the candidate’s keep/delete controls so useful words remain.
- Review the pending count and choose the panel’s Delete action when ready to apply those selections.
- Replay the affected joins. Keep the original file and export a separate result.
You can also drag-select a specific unwanted word or phrase and use Delete selection. The basic workflow is explained in editing video by editing text.
Understand the Smart edit toolbar shortcut
The top-level Smart edit shortcut is a different route from reviewing candidates in the Script panel. It runs silence and filler detection, and retake handling when transcript information is available, then adds changes to the current edit.
Do not assume the toolbar waits for a separate final confirmation list. After using it, inspect the timeline and play the result before export. Detection can cut something you meant to keep; retake matching does not understand which performance is better. Use the app control guide to choose the route you want. Advanced tools may require Pro; check current plans.
Removing filler words and removing silence are separate tasks
Filler words are audible speech. Removing them needs usable transcript timing. Silence detection instead looks for low-level audio intervals and can work without a transcript. A pause containing room noise may behave differently from a completely silent gap.
Cut long waits if they add nothing, but preserve the time needed to read a screen, hear an explanation or understand a transition. For a recording with frequent visual actions, check the picture as well as the sound around every proposed cut.
Check the edit, captions and final export
- Listen for clipped consonants, abrupt breaths and missing context.
- Check that a cut did not remove an important mouse action or screen change.
- If you need visible captions, proofread them and verify the burn-in option at export.
- Export using a name that preserves the source, then open the output in a player.
A subtitle correction alone does not remove footage. For caption corrections and SRT saving, see the offline subtitle workflow. For phone script reading rather than desktop post-production, use Penbeam Mobile.
Frequently asked questions
Can filler removal delete meaningful words?
Yes. Filler matching and transcript timing can be wrong, and words such as like or actually can be meaningful. Review context and playback before exporting.
Does silence removal need subtitles?
Silence detection works from audio levels. Spoken filler detection needs recognised words and usable timing, so generate the transcript first.
Does Smart edit wait for me to confirm every candidate?
The toolbar shortcut adds edits directly. The Script panel offers a separate review-first flow using Analyze and its pending deletion controls.
Download for your computer, check the transcript and export a short test. Review current plan limits before purchasing.