Echo & Reverb Remover (Fix Echoey Room Recordings)
Less room, more voice: echo and reverb turned down, with a before and after to compare.
Try before you buy.
- Free preview: the first half of the result (up to 30 seconds) at the quality you chose, with a short MySmartCoPilot chime every 10 seconds.
- Locked until you unlock it: download and share.
- Unlock: Pro pass, ₹179 for 30 days or ₹79 for 1 week, a one-time payment that never renews. Unlock: Pro pass, $5.99 for 30 days or $2.99 for 1 week, a one-time payment that never renews.
Ways to unlock shows how to get the full result.
Printing this result is locked in the free preview.
Echo removal
Higher removes more of the echo; very high can make voices sound thin. Listen to the before and after.
Before and after
Time runs left to right, pitch bottom to top, brighter is louder. Echo shows as the smear after each word; with it removed the words stand apart. Both versions play at the same loudness, so you compare the echo, not the volume.
Result
Locked in the free preview. Opens the ways to unlock this result.
Locked in the free preview. Batch runs unlock with a pass.
Locked in the free preview. Query results unlock with a pass.
About the Echo & Reverb Remover (Fix Echoey Room Recordings)
Speech recorded in an empty room, a classroom, a hall or on a phone at arm’s length rings on after every word: the echo of the room. This page turns that echo and reverb down in voice notes, lectures, podcasts, interviews and the sound of videos. It measures how long the room rings, lets you set the strength, and plays a before and after of the same twelve seconds at the same loudness, with a spectrogram of each, so you can hear and see what changed before you make the file. For a video only the sound is processed: the picture is copied as it is.
It is classic signal processing that runs in your browser on your own device — a delayed linear-prediction filter followed by a model of how a room’s echo dies away — so no AI model is downloaded and nothing is uploaded. It is an alternative to the online AI services that remove echo after you upload your audio.
The free preview on this page is the first half of the result (up to 30 seconds) with a short MySmartCoPilot chime every 10 seconds; for a video, the first half (up to 10 seconds) with a watermark. With a Pro pass, or after unlocking this one result, you get the whole file.
How to use it
- Choose or drop a recording or a video (MP3, M4A, WAV, FLAC, OGG, MP4, MOV, MKV, WebM…). The page measures its sound and how long the room’s echo lasts.
- Press Play under Before and after and switch between Original and Echo removed while it plays. Move Compare from to check another part.
- Set the Strength (higher removes more echo; too high can make a voice thin), and if the measured room seems wrong, pick the Room yourself. Tick the gentle noise gate if the background hiss between words bothers you.
- Choose the level of the result and the kind of file — the same kind as the original, WAV or MP3 — then press Remove the echo. Without a pass you get the free preview; with a Pro pass, or once this result is unlocked, Download or Share gives the whole file.
Examples
45-minute MP3 · measured echo about 0.7 s · strength 6 · same loudness as the original
lecture-echo-removed.mp3 at the same bitrate: the ringing after each sentence is much shorter, the voice is clearer, and the level matches the original.
MP4 (H.264 + AAC) · measured echo about 1.2 s · strength 8 · podcast level (−16 LUFS)
talk-echo-removed.mp4 with the picture copied unchanged and new AAC sound: noticeably less hall, at −16 LUFS with peaks under −1 dBTP. A very large hall never becomes a studio, but speech becomes easier to follow.
How it works
- Measuring the room. In four bands from 250 Hz to 4 kHz the page follows the sound after words end. Where it dies away freely, the slope between 10 and 30 dB below the peak gives the room’s reverberation time (RT60: how long the echo takes to fall by 60 dB). A dry recording measures 0.2 s or less and is hardly changed.
- Prediction. In each frequency band of the short-time spectrum (frames of 43 to 64 ms), a filter predicts the sound from what came at least 20 to 32 ms earlier, weighted by how loud the speech is, and takes what it can predict away. That is the weighted prediction error (WPE) method, ported from the open-source nara_wpe reference. The direct voice and its first reflections arrive inside that time, so they are kept.
- The echo that is left. One microphone cannot predict all of a room’s echo, so a second stage estimates the late reverberation from the measured RT60 — what was heard 50 ms earlier, decayed the way that room decays — and turns it down, never by more than the strength allows (6 to 20 dB).
- Blocks. Long files are processed in overlapping 30-second blocks that are cross-faded, on several of your device’s processor cores at once.
- Level. The result is brought to the original’s loudness or to −16 LUFS (ITU-R BS.1770 loudness), with a true-peak limiter at −1 dBTP; left as processed, it is only kept from going over full scale.
Choosing the strength
- 4 to 6 suit most voice notes, lectures and meetings recorded in ordinary rooms.
- 7 to 10 suit bare halls and stairwells; listen for a thin or “underwater” voice and step back if you hear it.
- 1 to 3 only take the edge off, for recordings that already sound almost right.
- The comparison plays both versions at the same loudness, so a version does not sound better just because it is louder.
Videos and formats
A video keeps its picture packet for packet (MP4, MOV, MKV and WebM); only the sound is processed and encoded again (AAC, or Opus in WebM). Audio files keep their kind — MP3 stays MP3 at a similar bitrate, WAV and FLAC stay lossless — or can be saved as WAV or MP3. A “stereo” file that is really one microphone is processed once, which is twice as fast.
Limitations
- Works best on speech with moderate echo. Very large halls and churches, music and several people talking at once improve less.
- It reduces echo; it does not remove traffic, hum or music in the background (the gate only lowers the quiet parts between words).
- Very high strength can make a voice sound thin or hollow; the before and after lets you check before you make the file.
- Up to 90 minutes on a computer and 15 minutes on a phone; cut longer recordings into parts first (MP3 Cutter, Video Trimmer). Phones take noticeably longer than computers.
- Files with several sound tracks keep only the main one, and 5.1 sound is mixed down to stereo.
Privacy
On this device; its files download once
Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot. The audio engine (and, when your file needs them, the MP3, AAC or FLAC encoder) downloads from MySmartCoPilot the first time you open a file. Your recordings and videos are never uploaded.
Frequently asked questions
What do I get without a pass?
Without a pass, Echo & Reverb Remover (Fix Echoey Room Recordings) shows the first half of the result (up to 30 seconds) at the quality you chose, with a short MySmartCoPilot chime every 10 seconds. Until you unlock it, the result can’t be downloaded or shared. A Pro, Premium or Ultimate pass, a one-time payment that never renews, unlocks the full result. The pricing page lists the passes and their prices.
Is my recording uploaded?
No. The file is read, processed and written again in your browser on your own device; nothing about it is sent anywhere.
Is this AI?
No. It is classic signal processing — weighted prediction error dereverberation and a statistical model of a room’s echo — so no AI model is downloaded, and the same file with the same settings always gives the same result.
How much echo can it remove?
It depends on the room and the strength. On a synthetic voice in simulated rooms with 0.4 to 1 second of echo, the default strength turned the echo between words down by about 5 to 8 dB and the strongest setting by about 7 to 12 dB, while the words themselves kept their level within about 1 to 4 dB. On your own recording the page measures the room first, and the before and after lets you hear the difference.
Why does the result sound a little quieter or drier?
The echo carried part of the energy of the original, so taking it out lowers the level a little; “Same loudness as the original” brings it back. A drier sound is the point: it is what the voice sounded like without the room.
Will the picture of my video change?
No. The picture is copied packet for packet into the new file; only the sound is processed and encoded again, so the video stays in sync and loses no quality.
What does the free preview include?
Without a pass the page processes the first half of your recording, up to 30 seconds, in the format you chose, with a short MySmartCoPilot chime every 10 seconds, and plays it here; for a video, the first half up to 10 seconds with a watermark. The whole file needs a Pro pass, or an unlock of this one result, which makes it from the file still open on the page.
Open-source components
This page can load these open-source components, each only when a task needs it:
- It uses FFmpeg under the LGPLv2.1 and the LAME MP3 encoder under the LGPL, through libav.js: licence and source code.
- It uses FFmpeg under the LGPLv2.1, through the Mediabunny add-ons: licence and source code.
- It uses the LAME MP3 encoder under the LGPL, through the Mediabunny MP3 add-on: licence and source code.
- It uses Mediabunny under the MPL 2.0: licence and source code.
You may replace them with your own builds. Their licences and the credits of every library the site uses are on the open-source credits page.