remove-video-silences

name	remove-video-silences
version	2.0.0
description	Remove silent sections from videos using Voice Activity Detection (VAD). Processes video in chunks, detects speech with Silero VAD, and stitches together speech segments while preserving natural flow. Use when asked to remove silence, dead air, or trim pauses from videos.
tags	["video","audio","silence","vad","ffmpeg","editing","remsi"]
author	Agent Zero

Remove Video Silences

Remove silent sections from video files using the elegant Remsi approach.

Based on Remsi - uses ffmpeg's silencedetect filter combined with select/aselect filters for clean, efficient processing.

Features

• Simple & elegant: Single ffmpeg command with select filters
• No dependencies: Only requires ffmpeg (no Python ML libraries)
• Preserves quality: Uses select filters, not re-encoding segments
• Configurable: Adjust silence threshold and minimum duration

Usage

python /a0/usr/skills/remove-video-silences/scripts/remove_silences.py INPUT OUTPUT [OPTIONS]

Options

• -n, --noise: Silence threshold in dB (default: -50)
• -d, --duration: Minimum silence duration in seconds (default: 0.9)
• --dry-run: Show ffmpeg command without executing

Examples

# Default settings (-50dB threshold, 0.9s minimum silence - conservative)
python scripts/remove_silences.py video.mp4 video_trimmed.mp4

# More aggressive silence detection
python scripts/remove_silences.py video.mp4 video_trimmed.mp4 -n -25 -d 0.3

# Preview command without executing
python scripts/remove_silences.py video.mp4 video_trimmed.mp4 --dry-run

How It Works

1. Detect silences: Uses ffmpeg's silencedetect filter to find silent sections
2. Build filter: Creates select='between(t,start,end)+...' expression for speech sections
3. Apply filter: Runs ffmpeg with matching video (select) and audio (aselect) filters
4. Fix timestamps: Uses setpts=N/FRAME_RATE/TB and asetpts=N/SR/TB to correct timing

Output Example

Processing: video.mp4
Parameters: noise=-30.0dB, min_silence=0.5s

Video duration: 54.00s
Silent sections found: 6
Total silence: 6.19s (11.5%)

Processing video...

==================================================
COMPLETE
==================================================
  Input:    54.00s
  Output:   47.83s
  Removed:  6.17s (11.4%)
  File:     video_trimmed.mp4

Threshold Guidelines

Default settings are conservative to avoid cutting voice. Adjust based on your needs:

Content Type	Recommended Threshold	Min Duration
Clean studio audio	-50 dB	0.9s
Normalized speech	-45 dB	0.7s
Noisy recording	-40 dB	0.5s
Aggressive cutting	-30 dB	0.3s

Credits

Based on the Remsi approach by bambax.