In this guide
A dynamics compressor reduces the difference between louder and quieter parts by turning down audio that exceeds a threshold. Use makeup gain cautiously, compare at similar loudness, and listen for lost consonants and raised background noise.
An audio compressor is a dynamics processor that reduces the level of louder moments when they exceed a chosen threshold. This can make a voice or instrument easier to hear consistently in a mix. Once the louder parts are controlled, makeup gain can raise the overall level, making quieter details more audible relative to the peaks. It can also make breaths and room noise more noticeable.
That meaning of compression is different from reducing file size. A dynamics compressor changes the sound's level over time; an MP3 encoder reduces data for storage. Saturalabs' audio file compressor makes an MP3 for delivery and does not provide the threshold and ratio controls described here.
Threshold and ratio, with an example
The threshold is the level at which an ordinary downward compressor begins to reduce gain. The ratio describes how much an input above that threshold is reduced. In an idealized hard-knee, steady-state example with a threshold of −20 dB and a 2:1 ratio, an input 6 dB above the threshold would emerge 3 dB above it: −14 dB in becomes −17 dB out, before makeup gain. Real compressors react over time, so a brief peak may behave differently.
A soft knee transitions into compression gradually around the threshold. The detector and metering also vary between processors; do not assume that a waveform peak predicts exactly how every compressor responds. See Audacity's compressor documentation for its controls.
What the main controls change
| Control | What it changes | Listen for |
|---|---|---|
| Threshold | Which louder passages trigger gain reduction. | Whether the processor reacts only to unusually loud words or to most of the performance. |
| Ratio | How strongly levels above the threshold are reduced once compression is established. | Whether speech feels controlled or unnaturally flattened. |
| Attack | How quickly reduction develops after the signal exceeds the threshold. | Changes to word beginnings and transient detail; slower attack can let peaks through. |
| Release | How quickly reduction eases after the signal falls. | Audible pumping or level changes between phrases. |
| Makeup gain | Raises the compressed signal after gain reduction. | Background sound, breaths and output peaks rising with the voice. |
Try compression on a voice recording
- Start with a copy and a goal. Choose a passage containing quiet words, loud words, breaths and room tone. Fix obvious clip-gain problems first if one isolated shout needs its own edit.
- Open a dynamics compressor. In a compatible version of Audacity, select the audio and choose Effect → Volume and Compression → Compressor. Control names and presets differ across versions; consult the current manual.
- Start gently. Try a modest ratio such as 2:1, then adjust the threshold while listening to the loud passage. Use the processor's gain-reduction display if available. A threshold value suitable for one recording may do nothing or too much on another.
- Adjust timing in context. Compare the starts and ends of words as you change attack and release. If the voice loses clarity or the background swells between phrases, ease the settings or try a different timing.
- Match levels for comparison. Use makeup gain carefully and compare processed and bypassed audio at similar perceived loudness. Louder can seem better even when detail is lost. Check the full export for clipping and noise before delivery.
Compressor, normalizer, limiter or gate?
A peak normalizer applies a fixed gain to reach a chosen sample peak; it does not separately reduce the loud words. A limiter controls peaks more aggressively, depending on its design; see the compressor-versus-limiter guide for choosing between them. A gate reduces sound below a threshold and can also suppress quiet words. The normalization guide explains peak and loudness targets, while the local noise gate lets you compare threshold-based reduction on a short file.
Compression is not noise removal or source separation. If music or another voice overlaps the speaker, investigate the cleanup-method guide before applying more gain. If you need a smaller attachment, use the file-size guide instead. Keep the original recording in either case.
Can a compressor fix distorted audio?
No compressor can reliably restore a waveform already clipped during recording. It can control later level changes, but a cleaner original take or a dedicated repair attempt may be needed. Check quiet consonants, breaths and room sound after processing; the best setting is the one that helps the intended mix without damaging the performance.