silenceremove

Remove silence from audio files by trimming quiet sections at the beginning, end, or throughout the audio.

Quick Start

Remove silence from beginning and end:

ffmpeg -i input.wav -af silenceremove=start_periods=1:start_duration=1:start_threshold=-60dB:stop_periods=1:stop_duration=1:stop_threshold=-60dB output.wav

Remove all silence throughout the file:

ffmpeg -i input.wav -af silenceremove=start_periods=1:start_duration=1:start_threshold=-60dB:stop_periods=-1:stop_duration=1:stop_threshold=-60dB output.wav

Core Concepts

What is Silence?

Silence is audio below a certain volume threshold for a minimum duration. You define:

• Threshold: How quiet audio must be to count as silence (e.g., -60dB)
• Duration: How long the quiet audio must last to be considered silence (e.g., 1 second)

Three Types of Silence Removal

1. Beginning Trimming: Remove silence from the start of the audio
2. End Trimming: Remove silence from the end of the audio
3. Throughout Trimming: Remove silence from anywhere in the audio (beginning, middle, and end)

Configuration Guide

Beginning Silence Removal

Controls what happens at the start of your audio file.

start_periods - How many silence-to-sound transitions to skip

• 0: Don't remove any beginning silence (default)
• 1: Remove the first stretch of silence at the beginning
• 2: Keep the first silence, remove the second silence, etc.

start_threshold - How quiet audio must be to count as silence

• Examples: -60dB, -40dB, 0.001 (linear scale)
• Quieter values (more negative dB) = stricter silence detection

start_duration - How long audio must be non-silent before stopping the trim

• Time in seconds: 0.5, 1, 2.5
• Once non-silence lasts this long, stop trimming from the beginning

start_mode - How to detect when silence ends

• any: Any channel above threshold counts as non-silence (default)
• all: All channels must be above threshold

End Silence Removal

Controls what happens at the end of your audio file.

stop_periods - How many sound-to-silence transitions to process

• 0: Don't remove any ending silence (default)
• 1: Remove the final stretch of silence at the end
• 2: Remove the last two silence sections
• -1: Special mode - Remove silence throughout the entire file (beginning, middle, and end)

stop_threshold - How quiet audio must be to count as silence

• Same format as start_threshold

stop_duration - How long silence must last before it gets removed

• Time in seconds
• Only silence lasting this long or longer will be trimmed

stop_mode - How to detect silence

• any: Any channel below threshold counts as silence (default)
• all: All channels must be below threshold

Advanced Options

detection - Which type of audio level to measure

• peak: Measure peak amplitude (default, most common)
• rms: Measure RMS (root mean square) level

window - Time window for RMS detection

• Only used when detection=rms
• Defaults to 0.02 seconds (20ms)

Common Usage Patterns

Clean Up Podcast Recording

Remove silence from beginning and end, keep natural pauses:

ffmpeg -i podcast.wav -af silenceremove=start_periods=1:start_duration=0.5:start_threshold=-50dB:stop_periods=1:stop_duration=2:stop_threshold=-50dB output.wav

Aggressive Silence Removal

Remove all silence throughout the file:

ffmpeg -i input.wav -af silenceremove=start_periods=1:start_duration=0.5:start_threshold=-40dB:stop_periods=-1:stop_duration=1:stop_threshold=-40dB output.wav

Conservative Cleanup

Only remove very obvious silence:

ffmpeg -i input.wav -af silenceremove=start_periods=1:start_duration=1:start_threshold=-70dB:stop_periods=1:stop_duration=3:stop_threshold=-70dB output.wav

Multiple Silence Sections

Skip the first quiet section but remove subsequent ones:

ffmpeg -i input.wav -af silenceremove=start_periods=2:start_duration=1:start_threshold=-60dB output.wav

Parameter Reference

Troubleshooting

Filter removes too much audio:

• Make thresholds more strict (more negative): -40dB → -60dB
• Increase duration requirements: 0.5 → 2
• Use detection=rms for more stable detection

Filter doesn't remove enough silence:

• Make thresholds less strict: -60dB → -40dB
• Decrease duration requirements: 2 → 0.5
• Check your audio levels with ffmpeg -i input.wav -af volumedetect -f null -

Removes silence in the middle when you don't want it to:

• Use stop_periods=1 instead of stop_periods=-1
• Only configure beginning/end trimming, not throughout-file trimming

Keeps silence you want to remove:

• Lower the stop_duration value
• Make sure stop_periods is set appropriately (1 for end-only, -1 for throughout)

Understanding the Algorithm

The filter works by:

1. Scanning audio for sections that meet your silence criteria (threshold + duration)
2. For beginning trim: removing silence until it finds non-silence lasting start_duration
3. For end trim: identifying silence sections and removing them based on stop_periods setting
4. The -1 mode for stop_periods enables a special algorithm that processes the entire file

This internal complexity is why the parameters can seem confusing - they're controlling different parts of this multi-step process.