You will learn#
- how silences are found and how the cut is made;
- which settings control it;
- how to remove silences with a command.
The idea#
When a video asset is imported, the product analyses its audio and records where the sound falls below a noise level. Silence removal replaces one clip with one clip per stretch of speech and, by default, closes the gaps so the later items of the track move earlier. Speech is never cut hard: a little padding stays on each side of every cut.
| Setting | Default | Meaning |
|---|---|---|
threshold_db | -35 | Noise level of silence: -30, -35, -40 or -45 |
min_silence | 0.4s | Silences shorter than this are kept |
padding | 0.1s | Speech kept around each cut |
ripple | track | track closes the gaps, none leaves them |
The command never starts an analysis itself: the service supplies the silences of the asset. Read them first with media_inspect and the section silences. Silence removal has no inspector section yet: use the command palette (Ctrl+K, Cmd+K on macOS), the CLI or an agent. The whole cut is one undo step.
Quick start#
vukem-video media:inspect ast_4h6j8k2m3n --project prj_7k2m9x4q1b --sections silences
vukem-video edit --project prj_7k2m9x4q1b clip.remove_silences item_id=itm_m1n2p3q4r5 \
min_silence=0.5s threshold_db:=-40media_inspect with {"project_id": "prj_7k2m9x4q1b", "asset_id": "ast_4h6j8k2m3n", "sections": ["silences"]}, then edit with:
{ "op": "clip.remove_silences", "args": { "item_id": "itm_m1n2p3q4r5", "min_silence": "0.5s" } }Use dry_run to see the pieces before you save, and view with a contact sheet to check the result.
Common mistakes#
- Using it on a clip of an asset with no audio: the command is refused.
- A threshold that is not one of the four values:
E_VALIDATION. - Expecting the captions of the clip to follow the cut. Re-sync them with
caption.retime_from_clip.