How to Remove Echo and Reverb from Audio

To remove echo and reverb from audio, start with the free Noise Reducer: for a speech-only recording, pick Voice mode, click Reduce noise, then compare the result against the original with the Original/Denoised slider and keep only as much as sounds natural. Be ready for a partial improvement, not a dry studio voice. Echo and reverb are among the hardest things to clean up, and no free tool, ours included, removes them completely. Everything runs in your browser, so your audio never leaves your device.

This guide explains why room sound is so stubborn, what you can realistically soften after the fact, and how to stop it before you press record.

Echo vs reverb: what you're actually hearing

Both come from the same thing: sound leaving the source, bouncing off surfaces, and reaching the mic late. The difference is timing and density.

Echo is a distinct repeat. The reflection arrives late enough, and stands out enough, that you hear it as a separate copy of the sound: "hello… hello". You get it from a far wall in a large space, or deliberately from a delay effect.

Reverb is thousands of reflections arriving so close together that they blur into one smooth, decaying wash. It has two parts:

• Early reflections are the first few bounces, off the desk, the nearest wall and the floor, arriving within roughly the first 50 milliseconds. Your ear doesn't hear them as separate events. They fuse with the voice and colour it, which is where that hollow, boxy, "recorded in a kitchen" quality comes from.
• The late tail is the dense decay that follows, the ring you hear after a word ends. The time it takes to fade by 60 dB is called RT60, the standard measure of how "live" a room is.

Flutter echo is a special case: two bare, parallel surfaces, like the walls of an empty room or a hallway, bounce sound back and forth in a rapid, metallic "zing" after a clap or a hard consonant.

If you're here because you want more space on a sound, that's the opposite job. See How to Add Reverb to Audio (Free, Online).

Why reverb is so hard to clean up

Most noise is a different signal from the voice. Hiss is random electrical noise; hum is a steady tone from the mains; a fan is a constant rush. None of them follows what the speaker says, so a tool can learn what the noise sounds like and subtract it. That's why How to Remove Hiss from Audio and the general approach to removing background noise from audio work as well as they do.

Reverb isn't like that. Every reflection is a delayed, filtered copy of the voice itself. It rises and falls with every syllable, shares the voice's pitch and spectrum, and changes as the speaker moves. There's no steady "noise profile" to learn, because the noise is the signal, arriving a little later.

Early reflections are the worst part: they overlap the direct sound so closely that pulling them apart without damaging the voice is extremely difficult. The late tail is a little easier, because it hangs in the gaps after words. That's why most cleanup, from any tool, softens the tail more than it fixes the boxiness. A roomy recording can usually be made less roomy; it rarely becomes a clean, close-miked voice.

What the Vocal Cut tools can and can't do

Voice mode (AI speech model). Voice mode runs GTCRN, a small AI speech-enhancement model that keeps the voice and suppresses what it judges not to be speech. The tool's own warning, aimed at people running songs through it, says it removes everything that isn't a voice, reverb tails included. On a speech recording that same behaviour may soften some of the diffuse room sound around the voice, especially the tail in the pauses. It isn't a dedicated de-reverb model, though, and results vary a lot by recording. It won't remove early reflections that have already blended into the voice. Try it and judge by ear. How AI Noise Reduction Works explains how models like this separate speech from everything else.

Music mode (spectral gating). Music mode isn't AI. It profiles the steady noise in your file and turns down frequency bands near that floor. Reverb isn't steady, so Music mode does little for it beyond trimming a bit of tail in quiet pauses. And don't run songs through Voice mode instead: it strips instruments, pads and cymbals along with the room.

Other tools. The Audio Enhancer is conventional processing, and its own FAQ says it cannot remove heavy room echo. The Vocal Remover splits a song into vocals and instrumental, but reverb on the vocal mostly travels with the vocal stem, so you get a reverby acapella, not a dry one.

For severe cases, dedicated de-reverb plugins and paid AI speech-repair services exist, and some do noticeably better than a general noise reducer. Even they work best on mild-to-moderate room sound. On a badly echoey take, re-recording is usually the real fix.

What you'll need

• A file in MP3, WAV, FLAC, M4A, AAC, OGG, Opus or WebM audio. Safari can't decode OGG or Opus, so open those in another browser.
• A file up to 100 MB and 12 minutes long. Trim a longer one first with the Audio Cutter.
• To remove echo from a video, first extract the audio from the video file you already have with Video to Audio, clean it, then put it back in your video editor.
• Headphones. Room sound is hard to judge on laptop speakers, which add their own.

How to remove echo and reverb from audio, step by step

1. Check what you're dealing with. Open the file in the Spectrogram. Reverb shows up as smeared trails extending to the right of each word into the gaps, rather than crisp edges. How to Read a Spectrogram shows how to read it. Long trails mean a lot of tail; a voice that looks clean but sounds hollow points to early reflections, which are harder to treat.
2. Open the Noise Reducer and add your file. Drag it onto the drop zone on the Noise Reducer page, or click to choose it.
3. Choose Voice mode. It's the default. Use it only if the recording is speech without music. The first run downloads the AI model (about 0.5 MB) once and then caches it.
4. Click Reduce noise (Voice mode). Processing runs in a background worker and shows a percentage, the current stage and an estimate of time left. It isn't instant; a long file takes a while.
5. Compare against the original. Press play. The slider runs from Original to Denoised and starts at 75%, shown as "Noise reduction: 75%". Drag it to 0% to hear the untouched recording, then back up. Listen to word endings and pauses, where the tail lives. Changes apply live and carry into the export.
6. Settle on the most natural blend. If the voice sounds thin, choppy or robotic, pull the slider back. Keeping some of the room is often better than a voice that sounds processed.
7. Export. Choose WAV (the default), MP3, or Opus where your browser supports it (the option is hidden where it isn't, for example in Safari), and click Export WAV (or the format you picked). Reset File clears everything to start again.

Voice mode processes the speech at a 16 kHz sample rate, so the cleaned signal is mono and limited to the speech band, roughly below 8 kHz. The export is mono too, with some of the full-range original blended back in at whatever slider position you choose. That's fine for podcasts, interviews and voice memos, and it's another reason it doesn't suit music.

Troubleshooting a roomy recording

It barely changed. That's common with reverb. If the room sound is mostly early reflections, the boxy colour baked into the voice, there's little a speech model can lift out. Try the EQ cut below, and consider re-recording.

The voice sounds watery or robotic, or word endings get clipped. The model is guessing hard where voice and reflections overlap. Pull the slider toward Original until the artifacts disappear.

There's also hiss, hum or fan noise. Voice mode handles those at the same time, and usually more reliably than the reverb. For a recording where background noise is the main problem, see How to Remove Background Noise from Audio.

It got quieter. Removing room sound removes energy. Run the cleaned file through the Volume Normalizer afterwards.

Reduce boxy room sound with EQ

Small, untreated rooms tend to build up energy in the low mids, which makes a voice sound boxy or boomy. The Equalizer has six fixed bands: an 80 Hz low shelf, peaking bands at 240 Hz, 750 Hz, 2.2 kHz and 6 kHz, and a 12 kHz high shelf, each ±12 dB.

Try a cut of a few dB at 240 Hz, and maybe a small cut at 80 Hz on a voice that doesn't need deep lows. Listen as you go. This doesn't remove reflections; it turns down the frequency range where a small room's buildup tends to pile up, so the voice sounds less like it's in a box. It's a blunt tool: cut too much and the voice loses warmth and body. Parametric vs Graphic EQ: The Difference explains why fixed bands can't target a specific room resonance precisely.

When to re-record instead

Re-record if you can hear distinct repeats or flutter on hard consonants, if the voice still sounds distant with the tail softened, or if cleanup makes it sound processed before the room sounds acceptable. For a short take like a voiceover line or intro, five minutes in a better spot beats an hour of repair. Save heavy cleanup for audio you genuinely can't capture again, like an interview or an event recording.

Prevent echo and reverb at the source

Room sound is far easier to avoid than to remove.

• Get the mic closer. This is the biggest single fix. Halve the distance and the direct voice gets louder relative to the room, so the reflections matter less. A few inches to a hand's width from the mouth is typical for speech. Mic Placement for Clean Voice Recordings covers distance, angle and plosives in detail.
• Use soft furnishings. Heavy blankets, duvets, curtains and cushions soak up reflections. A thick blanket hung behind you, facing the mic, catches the reflections bouncing off the wall at your back. A closet full of hanging clothes is a classic home voiceover booth.
• Keep off-axis from hard walls. Don't face a bare wall or window at close range, and avoid empty rooms and bathrooms, whose tiles, glass and parallel walls cause flutter echo.
• Do a test first. Record a few seconds with the Audio Recorder (or the 5-second sample in the Mic Test), clap once, and listen for ringing.

For a full setup walkthrough, see Record a Voiceover With an Online Voice Recorder. More cleanup and recording guides are in the Audio How-To Guides.

Frequently asked questions

Can you completely remove echo and reverb from a recording?

Usually not. Reverb is made of delayed copies of the voice itself, so it overlaps the sound you want to keep in both time and frequency. Tools can often soften the tail that rings into pauses, but early reflections blend into the voice and are very hard to separate cleanly. Expect a noticeable improvement on mild room sound and a partial one on heavy echo, where re-recording is often the better fix.

What is the difference between echo and reverb?

Echo is a distinct repeat you can hear as a separate copy of the sound, caused by a reflection arriving late enough to stand apart. Reverb is many reflections arriving so close together that they merge into one decaying wash. Reverb has early reflections, which colour the voice, and a late tail, which rings on after each word. Flutter echo is rapid repeats between bare parallel walls.

Does the Vocal Cut Noise Reducer remove reverb?

It isn't a dedicated de-reverb tool. Voice mode uses an AI speech model that keeps the voice and suppresses other sound, and on speech recordings it may soften some of the room tail around the voice. Results vary by recording, and it can't remove reflections already blended into the voice. Music mode targets steady noise like hiss and hum, so it does little for reverb. Compare with the Original/Denoised slider and judge by ear.

Why does my voice recording sound echoey?

The microphone is picking up sound bouncing off hard surfaces in the room, such as bare walls, windows, desks and floors. The farther the mic is from your mouth, the louder those reflections are compared with your voice. Empty rooms, bathrooms and kitchens are the worst. Moving closer to the mic and adding blankets, curtains or cushions around you reduces echo far more than any cleanup afterwards.

Can I remove reverb from a vocal in a song?

Not well with free tools. Splitting a song with a vocal remover isolates the vocal, but its reverb mostly comes along with it, so you get a reverby acapella. Running that through a speech model would also strip musical detail and return mono audio. Dedicated de-reverb plugins can help with mild cases. If you have the original dry recording or stems, those are a far better starting point.