In this guide
Start with a moderate ratio and adjust the threshold to the recording, not a universal preset. Compare at a similar loudness, watch gain reduction, and check quiet words, noise and peaks before export.
For spoken voice in Audacity, start with a modest compressor ratio such as 2:1, then adjust the threshold while listening to a representative phrase. Set make-up gain carefully, compare the processed and original audio at a similar listening level, and check the exported file for clipping and unnatural pumping. There is no single best threshold: it depends on the level of the recording.
A compressor reduces the difference between louder and quieter parts; it is not a noise remover or a repair for an already distorted recording. Keep an untouched original and test a short excerpt with a quiet sentence, a loud word and a pause.
Where to find the Audacity compressor
For the current compressor, select audio and choose Effect → Volume and Compression → Compressor. You can also add Compressor as a realtime track effect to adjust and bypass it during playback. Audacity's realtime compressor guide describes the effects panel and its input, output and compression meters. Realtime processing is included in an export.
Controls differ in older releases and in the separate Legacy Compressor. The settings below describe the modern compressor; consult the Audacity compressor reference for your version rather than copying values into a different interface.
What each setting does
| Control | What it changes | What to listen for |
|---|---|---|
| Threshold | The level around which compression begins. Lowering it generally makes more of the voice trigger compression. | Words should become more even without flattening every phrase. |
| Ratio | How strongly the level above the threshold is reduced. At 2:1, an ideal steady signal 4 dB above threshold is reduced to about 2 dB above it, before make-up gain and knee effects. | Higher ratios can sound more controlled but also less natural. |
| Attack and release | How quickly compression takes hold and relaxes. Audacity's modern controls use milliseconds. | Too much attack shaping can dull consonants; an unsuitable release can make background level pulse. |
| Knee width | How gradually compression begins around the threshold. | A softer transition may sound less abrupt; compare rather than assuming it fixes every source. |
| Make-up gain | A fixed level change after compression. | Added gain can raise noise and peaks. Watch output levels and match loudness before comparing. |
A starting workflow for spoken voice
- Prepare the recording. Preserve the original, fix obvious editing problems and listen for clipping or strong noise. Compression will not reconstruct clipped syllables and can make background noise more noticeable after make-up gain.
- Use a mild starting point. In the modern effect, try a 2:1 ratio and a modest knee. Leave attack and release at the version's starting values initially. Audacity also offers factory presets; use one as a starting point, not proof that the result is right.
- Set the threshold by the actual signal. With realtime metering visible, play both quiet and loud phrases. Lower the threshold until the louder phrases show some gain reduction, then listen. If nearly everything is being compressed or the voice sounds flat, raise it or reduce the ratio.
- Adjust make-up gain deliberately. The modern Audacity manual lists a nonzero default make-up gain, so check it rather than assuming bypass and processed audio are equally loud. Try zero for a level comparison, then add only what the project needs without clipping.
- Compare and export. Toggle the effect while replaying the same passage at a similar perceived level. Listen to word beginnings, breaths, pauses and the loudest sentence. Export a WAV working copy and reopen it to check the actual file.
A threshold of −18 dB is only an example, not a recommendation for every recording: a quietly recorded voice may rarely cross it, while a loud one may trigger compression almost continuously. The meter and your ears are more useful than a copied number. Audacity's compressor and limiter guidance explains presets and realtime metering.
Troubleshoot the result
- Nothing seems to change: confirm that the effect is enabled on the track being played, and check whether the signal reaches the threshold. Temporarily lower it to hear the difference, then readjust.
- Breaths and background are too prominent: reduce the amount of compression or make-up gain. If the original contains steady hiss, compare a separate, restrained noise-reduction workflow.
- The voice pumps or sounds dull: back off the threshold or ratio and compare attack and release changes one at a time. Listen across phrases, not just a single loud peak.
- The export clips: reduce gain and check the full output again. A quieter export is preferable to distortion; peak readings and perceived loudness answer different questions.
For a short mono or stereo MP3/WAV, the Saturalabs loudness meter can measure integrated loudness and peaks locally after export. Inputs are limited to five minutes and 50 MiB; codec and measurement limitations are described on the tool page. A measurement supports the listening check, but does not by itself establish that the voice sounds natural.
Does compression remove noise or increase file compression?
No. Dynamics compression changes the relationship between louder and quieter sound; it does not identify noise or automatically shrink the file. MP3 encoding reduces file data by a different process. Our audio compressor guide explains the distinction, and compressor versus limiter compares peak-control roles.