Audio compression ratio explained

A compressor only acts on signal that crosses its threshold. Ratio decides what happens to that excess. If you are new to the wider subject, start with what compression does in audio and come back to this.

Read a ratio as input to output. At 4:1, every 4 dB by which the signal exceeds the threshold is allowed out as 1 dB. The threshold itself never moves. Only the material above it is scaled.

So a signal arriving 12 dB above the threshold leaves 6 dB above at 2:1, 3 dB above at 4:1, and a little over 1 dB above at 10:1. The higher the first number, the less of your original dynamic range survives.

Between 1.5:1 and 3:1 the effect is corrective. You are nudging a voice into a steadier place and a listener will not notice the processing, only that the delivery sits still.

Between 4:1 and 8:1 the compressor becomes an audible part of the sound. This range is useful for a voice that swings widely in level, or for a recording made in a room where you could not control how close the speaker stayed to the microphone.

Above 10:1 you have left compression behind and moved into limiting territory. Very little gets past the threshold. Used deliberately at the end of a chain this is protective; used across a whole voice track it flattens the performance.

A 10:1 ratio with a threshold set so high that your voice barely reaches it will do almost nothing. A 2:1 ratio with a threshold set well below your quietest word will squash everything you recorded. The two controls only make sense together.

The practical way to set them is to fix the ratio first at something modest, then lower the threshold slowly while watching the gain reduction meter. Stop when the meter shows a few decibels of movement on your louder phrases and returns to rest between them. Only raise the ratio if that amount of reduction is not enough to hold the voice steady.

Most conversational recordings are well served between 2:1 and 3:1. That is enough to stop a laugh or an emphatic sentence from jumping out of the mix, while leaving the natural rise and fall that makes someone sound like a person rather than an announcement.

Go higher when the source demands it. A remote guest on a laptop microphone who leans in and out of frame may need 4:1 or more simply to stay intelligible. An interview recorded on a lapel microphone at a fixed distance rarely needs anything above 3:1.

If you would rather hear the result than reason about the numbers, run the file through an audio compressor and compare it against the original.

A limiter is a compressor with the ratio turned up so far that no amount of signal above the threshold produces any meaningful increase at the output. Written as a ratio that is infinity to one, and in practice anything above about 20:1 behaves this way.

That makes a limiter the right tool for one job: stopping peaks from reaching the ceiling and clipping. It is the wrong tool for shaping how a voice sits across a whole episode, which is what the gentler ratios are for.