Audacity remove echo: what can be fixed and what cannot

When you speak in a hard room, your microphone captures your voice plus thousands of reflections of it arriving fractions of a second later. Those reflections are mixed into the same signal as the direct sound, at the same frequencies.

Separating them afterwards means working out which parts of a waveform were the original and which were the room, which is a much harder problem than removing steady background noise. Noise reduction works because hiss has a predictable pattern. Reverb is your own voice, delayed.

Audacity has no de-reverb effect, and any tutorial claiming a particular setting removes echo is describing something that reduces the impression of it while damaging the voice.

That is worth knowing before you spend an evening on it. The rest of this page is about what helps and by how much.

There is a completely different kind of echo worth ruling out first: the recording contains two copies of the same audio, slightly offset, because two things were recording at once or a track was duplicated by accident.

This sounds like a very clear doubling rather than a room. Look at your track list: if there are two tracks with near-identical waveforms, delete one and the echo is gone.

It also happens on a video call recorded at both ends, or when a participant's speakers fed back into their microphone. If the doubling only affects one speaker, that is what happened.

A noise gate helps most, because room echo is most audible in the tail after each phrase. Gating closes the track during pauses, which cuts the ring-out. Set a gentle reduction rather than full silence, and a decay around 200 milliseconds so words do not get clipped.

EQ helps a little. Room reflections tend to build up in the low mids, so a modest cut around 200 to 400 Hz can reduce the boxy quality. Keep it small, because that range also carries the body of the voice.

Noise Reduction with a profile taken from a pause can shave a little off, but the profile will include your reverb tail and so will start eating consonants quickly. Keep the reduction under about 6 dB if you try it.

Together these make a bad recording somewhat better. None of them make it sound like it was recorded in a treated room, and expecting that is how people end up with a processed, underwater result that is worse than the original.

Get closer to the microphone. This is the single most effective change and it costs nothing. At six inches your voice is far louder relative to the room than at two feet, so the reflections sit much lower in the mix.

Put soft things in the room. A rug, curtains, a sofa, coats, a duvet on a stand behind you. Reflections come off hard flat surfaces, and breaking those up matters more than any product you can buy.

Avoid recording in the middle of a room, and avoid facing a bare wall. Speaking into a corner with clothes hanging in it is the classic improvised solution, and it works better than it sounds like it should.

A dynamic microphone picks up less room than a condenser, which is why broadcast studios use them. If you are recording in a spare bedroom, that choice does more than any processing.

If the echo is mild and the content is good, a gentle gate and a small EQ cut, then leave it. Listeners tolerate a slightly roomy recording far better than an obviously processed one.

If it is severe and the recording is repeatable, record it again in a better spot. Twenty minutes of re-recording beats three hours of processing that ends somewhere worse.

If it is severe and cannot be repeated, an interview, a live event, a one-off, do the gentle treatment and publish it. An echoey recording of something worth hearing is worth hearing. A destroyed one is not.

Hilite cleans noise and levels automatically but does not claim to remove room reverb, because nothing reliably does. We make it, so weigh that as you like. The honest advice for echo is always about the room.