Audio Compressor (Dynamics)
Tame a recording where the volume keeps jumping around.
or paste one from your clipboard
How to use it
- Drop the recording onto the page. The transport plays the compressed result, so every change you make is audible immediately against the file you started with.
- Start from a preset. Voice and podcast suits anything spoken; gentle is the least destructive place to begin on music.
- Move the threshold until the loud moments are being caught and the quiet ones are left alone, then set the ratio to decide how firmly they are held down.
- Shape the character with attack and release: a slower attack keeps the punch of consonants and transients, a longer release stops the level pumping between words.
- Raise the makeup gain to bring the whole file back up to a usable level, then choose a format and download.
What a compressor is actually doing
A compressor is an automatic volume control with a rule. It watches the level of the signal, and whenever that level rises above the threshold it turns the signal down. Not all the way down, only by the fraction set by theratio. At 4:1, four decibels of input above the threshold come out as one. Below the threshold nothing happens at all, so quiet passages are passed through exactly as they arrived.
The consequence is that the distance between the loudest and quietest parts of the recording shrinks. That is the entire point. A file where someone leans into the microphone and then drifts away is unusable not because it is too loud or too quiet but because it is both, in the same file. Pull the peaks down, then usemakeup gain to lift the whole thing back up, and the quiet parts finish louder than they started while the loud parts land roughly where they were.
Attack and release, the two people get wrong
Attack is how long the compressor waits before clamping down once the signal crosses the threshold. Release is how long it takes to let go afterwards. Both are measured in milliseconds, and both change the character of a recording far more than the ratio does.
A very fast attack of one to three milliseconds catches the initial transient of every consonant, drum hit and pick stroke. That transient is where most of the perceived clarity lives, which is why heavily compressed material sounds soft and far away. Setting the attack to 15–30 ms lets the transient escape and compresses only the body of the sound behind it, which controls the level while keeping the punch.
Release is where the audible artefacts come from. Too short, and the compressor fully recovers during every pause, so the room noise under the voice swells and drops with each syllable. The effect usually described as pumping or breathing. Too long, and it never recovers, so a single loud moment holds everything after it down for seconds. For speech, 100–250 ms is a dependable range. For music, a release roughly in step with the tempo stops the compressor fighting the groove.
When not to compress
If the only problem is that a file is quiet from end to end, compression is the wrong tool: that is a gain or loudness-normalization job and it costs you nothing in dynamics. If a recording is dynamic on purpose. A classical performance, a sound effect, anything where the contrast is the content. Compressing it removes the thing that made it work. And if you are delivering to a streaming platform, they normalize loudness at playback, so compressing harder to be louder does not make you louder. It makes you flatter, at the same volume as everybody else.