What a compressor is for
Dynamic range is the gap between the quietest and loudest parts of a recording. When that gap is too wide you get the familiar problem of having to turn the volume up to hear the quiet bits and then being startled by the loud ones. A compressor narrows the gap: above a threshold you set, it reduces gain, so peaks come down and the whole file can then be raised without those peaks hitting the ceiling.
Nothing about this is the same as the compression in an MP3. Despite the shared word, a dynamic range compressor changes how loud things are relative to each other; MP3 compression discards data to make a file smaller. They are unrelated operations that happen to share a name.
The four controls
- Threshold, from -100 to 0 dB — the level above which reduction starts. Nothing below it is touched. Around -18 to -24 dB is a normal starting point for speech.
- Ratio, from 1 to 100 — how hard the reduction is. At 4:1, a signal 8 dB over the threshold is reduced to 2 dB over. 2:1 to 4:1 is gentle and natural; above 10:1 you are limiting rather than compressing.
- Attack, 0 to 1 second — how quickly reduction engages. Fast attacks catch transients and can dull consonants; slower attacks let the initial impact through, which usually sounds livelier.
- Release, 0 to 5 seconds — how quickly it lets go. Too short produces audible pumping as the level breathes up and down; 0.1 to 0.3 seconds suits most speech.
Tip: If compression is audible as an effect, the ratio is too high or the release too short. Good compression on speech is noticed only by its absence — the recording simply stops having loud and quiet patches.
Typical starting points
For podcast or interview speech, try a threshold around -20 dB with a 3:1 ratio, a fairly quick attack and a release near 0.2 seconds, then raise the overall level afterwards with the Volume Changer. For music, go gentler: 2:1 at a higher threshold, with a slower attack so drum transients survive.
Order matters in a clean-up chain. Reduce noise before compressing, because a compressor raises quiet passages and will lift any remaining hiss along with the voice. Trim, then reduce noise, then compress, then set final volume. Processing runs on the server at export and the uploaded file is deleted straight afterwards.