Trying to remove a horrible echo

First, I would just like to thank anyone who decides to view this forum post.

Our issue, as the subject line states, is a horrible echo. The context of the audio is that we recorded a Zoom Meeting via OBS and one of the input devices was on the same track as another, therefore created a very blatant echo. Since this was a one-time event, we are unable to “re-do” the meeting and are currently trying to salvage the audio that was captured and make it bearable. The one thing that makes the echo in our audio stand out from other “echo reducing” audio clips is that our echo is very voice-based instead of a general room-echo (see attached audio clip). I have tried a great many things from different forum posts and articles to resolve the echo but to no avail. Some of these include: creating noise profiles and applying different reduction, sensitivity, and smoothing levels based on the profile, using a high-pass filter, adjusting graphic eq, as well as using the Noise Gate plugin (mostly followed this guide,audio%20gets%20cut%2C%20reduce%20it.).

If anyone has any ideas about reducing the strength of the echo itself or any other tips, we would appreciate anything at all.
Thanks again!

Unfortunately that audio is beyond saving. It is sometimes possible to reduce reverberation, but there’s no way to remove strong echoes like that (other than perhaps some bleeding edge experimental use of TensorFlow, but that’s way beyond the scope of this forum).

I was truly thinking the same thing, just sadly don’t want to admit it. Thank you for your insight and quick response.

It’s not a classic echo because the main sound and the secondary sound are almost exactly the same. I found a sound at about 42 seconds where the desired show sound was brief and alone. The second blob is the damage nearly exactly 0.2 seconds later.

Screen Shot 2020-12-10 at 8.29.52 AM.png
Note they’re the same loudness which takes care of Noise Gate. Noise Gate needs a volume difference.

The two sounds are very, very nearly identical, so that shoots equalizers and filters in the head. They need a tonal difference.

It’s not classic echo because it doesn’t gradually die out over time. It’s one and done.

I got interested because fuzzy theory has it that knowing the delay, it should be possible to manually construct an “anti-echo.” It didn’t work. I could affect the real-time damage only by creating another damage sound 0.2 seconds later than that.

Screen Shot 2020-12-10 at 8.46.52 AM.png
So the middle cancels and I moved the damage from 0.2 seconds to 0.4. We’re not gaining on it.

I think we defined the problem wrong. We’re not getting rid of an echo as much as we are trying to separate individual instruments or voices in a band. We can’t do that, either.

Everybody knows you can’t record a Zoom call. How are you doing it? We’ve been recommending letting Zoom do it. You are the first posting that said, in effect: this normally works perfectly but we had this one screw-up. How does it normally work perfectly? Are you following someone’s instructions? A web posting? YouTube instructions?


Just want to say thanks to Koz for the in-depth analysis and response, it is much appreciated. Also, we talked with our speaker and are going to re-record it (go lucky I guess), therefore there may be no more need to research this issue.

But to respond to Koz’s questions: Because of our current AV setup, we were broadcasting a zoom over projectors for a group of students at a school, while having the ability to communicate with a mic through the zoom and our AV speakers at the same time. Because of this, we decided to record the entire session in OBS, with our soundboard running through VoiceMeeter. But some audio settings were assumed to be correct before the recording took place, but they were not therefore inevitably causing the echo. Our AV setup is currently super “wonky” and is band-aided right now because some parts have failed on us, but we continued in an effort to do the best with what we had.

Sorry for getting off-topic from Audacity and reducing the echo from the clip, but I also want to thank everyone who tried to assist us. Your analysis was able to get me a concrete answer and I appreciate the confirmed inability to easily reduce the echo. You all are the best!

Sorry for getting off-topic from Audacity

That’s a semi-joke. My “thing” is before and after Audacity. Not during. Its hard to wander too far off, but if you do, we’ll tell you.

Our AV setup is currently super “wonky” and is band-aided right now because some parts have failed on us

Yup. That’s normal. It’s mind-bleedingly complex and you’re up against the scientists and developers at Zoom doing everything in their power to thwart you. That’s why our go-to recommendation is let Zoom record it.

They are crystal clear how hard local recording is and as near as we can tell, have made Zoom system recording as comfortable as possible. Fair warning, we are told that due to the sickness, Zoom conferences have become insanely popular and preparing and shipping conference recordings to you can take some time.

There are local recording tricks if you can throw machines at it. Connect a fully qualified Zoom machine with no microphone or camera. Record the speaker feed. It will have all the other people on it.

we talked with our speaker and are going to re-record it

I’m not shocked. Sometimes if you beg enough…


Thanks for dropping back in.

The more hands-on information we can get, the faster we can get forum posters out of trouble.