Silence Remover
A threshold nearer zero judges more strictly. Use -50 for quiet recordings and -40 or -30 when there is room noise.
Add files and press start. Everything runs locally in your browser; nothing is uploaded
How to Use
📖 Tool Introduction
Strip silent passages out of audio automatically, entirely in your browser. Screen recordings, meeting captures and spoken podcasts are full of pauses where nobody is talking, and cutting them by hand takes ages. This handles them in one pass.
Two settings work together. The threshold decides how quiet counts as silence, and the minimum duration decides how long a gap has to be before it goes. Both have to hold: only stretches longer than the duration and quieter than the threshold are removed, so short breaths survive.
Threshold is in decibels and the closer to zero the stricter it is. A quiet room recording is read correctly at minus 50; with air conditioning or keyboard noise in the background, loosen it to minus 40 or even minus 30, otherwise the room tone counts as content and nothing gets cut. Half a second suits most material, while a lecture recording can go to one or two seconds to remove only obvious pauses.
A small margin of the original sound is kept around every cut so word beginnings and endings are not clipped, which would sound abrupt. That margin is fixed, so a listen-through afterwards is worth the time, especially with fast speech.
Processing uses a WebAssembly build of FFmpeg: about 24 MB downloads the first time and is cached after. Audio is never uploaded and stays on your device.
One more thing worth knowing before you start: cutting on a keyframe boundary is what makes the fast path possible, and those boundaries sit every second or two in most encoded video. That is why a requested start time can land slightly early or late with frame accuracy off.
One more thing worth knowing before you start: removing silence changes pacing as well as length. Cutting every pause makes speech feel breathless, so leaving the short ones alone is usually the better edit even when the tool could remove them.
✨ Key Features
- Automatic detection using both a loudness threshold and a minimum duration, so breaths are safe
- Four thresholds from -30 to -60 dB for different room noise
- Four durations from 0.3 to 2 seconds controlling how aggressively cuts are made
- A margin around each cut so word edges are not clipped
- Five output choices: keep source, MP3, WAV, FLAC, M4A
- Batch mode: add several files and each is processed in turn
🚀 How to Use
- Drag audio into the drop area, or click to choose files
- Pick a threshold: -50 for quiet recordings, -40 or -30 when there is room noise
- Pick the minimum silence length, 0.5 s for most material and 1 s for lectures, then either keep the source format or choose another
- Press start and watch the progress bar
- Listen through afterwards to confirm no word edges were clipped
⚠️ Things to Note
- Engine size: The first run downloads about 24 MB of WebAssembly engine, cached for later visits
- Choosing a threshold: Use -50 dB for quiet recordings. With air conditioning or keyboard noise, loosen to -40 or -30, otherwise room tone counts as content
- Choosing a duration: A longer minimum is more conservative in both directions. At 0.3 s even short pauses go, which suits fast speech but can sound rushed; at 2 s only obvious gaps go, which suits lectures
- Listen afterwards: The cut margin is fixed, so fast speech can lose a word edge. A quick listen is worth it
- Privacy: Audio is processed locally and never uploaded. Reloading the page clears it completely